{"status":"ok","feed":{"url":"https://newsletter.kiin.bio/feed","title":"Kiin Bio Weekly","link":"https://newsletter.kiin.bio/","author":"Kiin Bio","description":"Where AI meets Life Science","image":"https://substackcdn.com/image/fetch/$s_!UiEF!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8654bb64-0b90-4220-9c12-7c9269dd2c95_1093x1093.png"},"items":[{"title":"Vilya Research's Vilya-1, TU Munich's TWIN, and Stanford's UCE","pubDate":"2026-07-16 17:01:34","link":"https://newsletter.kiin.bio/p/vilya-researchs-vilya-1-tu-munichs","guid":"https://newsletter.kiin.bio/p/vilya-researchs-vilya-1-tu-munichs","author":"Natasha Kilroy","thumbnail":"","description":"Kiin Bio's Weekly Insights","content":"\n
Welcome back to your weekly dose of AI news for Life Science! This weeks fix:
\n
\n
Vilya-1 predicts macrocycle conformations across diverse chemistries with a single all-atom model, beating physics-based methods and existing deep learning approaches on geometric accuracy. Macrocycles are notoriously hard to model, and this is the first foundation model built specifically for them.
\n
TWIN is an implicit solvent ML potential trained on ab initio data that runs two orders of magnitude faster than explicit solvent while approaching DFT accuracy. If this holds up broadly, it removes one of the biggest bottlenecks in drug and peptide simulation.
\n
UCE is now published in Nature after a 2023 preprint. A single-cell foundation model that embeds any cell type without retraining, trained on the Tabula Sapiens atlas. The zero-shot transfer results across tissues and species are what earned it the Nature publication.
\n
\n
\n
Kiin Pioneer Programme
\n
We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.
\n
The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
Macrocyclic peptides occupy a sweet spot in drug design: large enough to hit protein-protein interactions that small molecules cannot reach, small enough to potentially cross cell membranes. The problem is that their conformational behaviour is extremely difficult to predict. They are too large for small-molecule force fields to handle well, too flexible for protein structure prediction methods, and too chemically diverse (cyclic, branched, non-natural amino acids, N-methylation) for any single modelling approach to cover.
\n
Current options are either expensive physics-based sampling (molecular dynamics, metadynamics) that takes days per compound, or deep learning methods built for linear peptides that do not generalise to the cyclic, heavily modified structures that matter therapeutically. Tools like AlphaFold and OpenFold were not designed for this chemical space. Vilya-1 from Vilya Research is the first foundation model purpose-built for macrocycle structure prediction, using a unified all-atom representation that handles diverse topologies and chemical modifications in a single model.
Predicting macrocycle 3D structure is hard because they span a chemical space between small molecules and proteins that existing methods handle poorly. Sturmfels, Salem, Hiranuma et al. from Vilya Research present Vilya-1, a deep learning foundation model trained on heterogeneous structural datasets across diverse macrocycle topologies and chemistries.
\n
Uses a uniform all-atom representation that accommodates cyclic peptides, non-natural amino acids, N-methylated residues, and other modifications without needing separate model architectures for each. Trained on a mix of experimental and computational structural data.
\n
Improved geometric accuracy compared to physics-based methods and existing deep learning alternatives. Coverage extends to small molecules. Also supports predicting developability properties like membrane permeability, and generative design of novel macrocycles with specified properties.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
Macrocycles are having a moment in pharma. Peptide drugs like semaglutide proved the market, and companies like Bicycle Therapeutics are building pipelines around constrained cyclic peptides. The structural prediction gap has been a real bottleneck: you can design a macrocycle computationally, but confirming it folds the way you expect requires either NMR (expensive, slow) or crystal structures (often impossible for flexible molecules). A model that gives you reliable conformational predictions for chemically diverse macrocycles would accelerate the design-make-test cycle considerably. The extension to permeability prediction is smart, since oral bioavailability is the other major hurdle for this drug class. No public code yet, which limits immediate evaluation by the community.
Molecular dynamics simulations of drugs and proteins in solution face a fundamental trade-off: explicit solvent (modelling every water molecule) is accurate but extremely expensive computationally, while implicit solvent models (treating water as a continuum) are fast but sacrifice accuracy. For drug design, this means you either spend days simulating a single compound in explicit water, or you get fast results from implicit models that often disagree with experiment.
\n
Previous implicit solvent approaches (GB/SA, COSMO) are based on empirical physics approximations that break down for many drug-like molecules. Recent ML potentials (ANI, MACE) have improved accuracy for gas-phase simulations, but extending them to solvation has been limited. TWIN from TU Munich takes an equivariant graph neural network, trains it exclusively on ab initio and experimental solvation data (no empirical force field data), and produces an implicit solvent potential that transfers across drugs, peptides, and proteins.
Simulating molecules in water is either accurate and slow (explicit solvent) or fast and unreliable (implicit solvent). Eckwert and Zavadlav from TU Munich present TWIN (Transferable Water Implicit Network), an ML potential that combines the speed of implicit solvent with accuracy approaching DFT-level explicit solvent calculations.
\n
Built on an equivariant graph neural network trained exclusively on ab initio quantum mechanical data and experimental measurements. Avoids any dependence on empirical force field parameters, which is what limits the transferability of traditional implicit solvent models.
\n
Two orders of magnitude faster timestep evaluation than explicit solvent approaches. Outperforms prior ML-based implicit solvent models on crystallographic and NMR benchmarks. Transfers across drug molecules, peptides, and proteins without retraining.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
Solvent is the hidden cost of computational drug design. Every binding free energy calculation, every conformational sampling run, every MD simulation spends most of its compute on water molecules rather than the drug you care about. If TWIN delivers on the promise of DFT-level solvation accuracy at implicit-solvent speed, and transfers reliably across chemical space, it could make free energy perturbation and metadynamics calculations accessible at scale rather than as expensive one-off studies. The key caveat: \u201capproaching ab initio accuracy\u201d on benchmarks does not always translate to \u201caccurate enough for rank-ordering drug candidates.\u201d The real test is whether medicinal chemists can trust it for prospective predictions. Two orders of magnitude speedup is the kind of change that alters what questions you can afford to ask.
Single-cell foundation models have proliferated over the past two years: scGPT, Geneformer, scFoundation, and others all aim to learn general representations of cells from large-scale scRNA-seq data. The shared limitation is that most require fine-tuning on task-specific labelled data, and they struggle to generalise to cell types or tissues not seen during training. They also typically use gene-token vocabularies tied to a specific species, making cross-species transfer difficult.
\n
UCE from Jure Leskovec and Stephen Quake\u2019s groups at Stanford takes a different approach: it embeds cells using protein-level information (via ESM embeddings of gene products) rather than gene-name tokens, which makes the vocabulary species-agnostic. Trained on 36 million cells from the Tabula Sapiens atlas across multiple human tissues. The preprint appeared in late 2023, and the Nature publication confirms the approach held up through peer review with additional validation.
Single-cell foundation models require fine-tuning for new tasks and do not transfer well across species or unseen cell types. Rosen, Roohani, Agrawal et al. from Stanford present UCE (Universal Cell Embedding), a foundation model that generates cell embeddings in a shared space without any task-specific retraining.
\n
Uses ESM protein language model embeddings to represent genes (rather than arbitrary gene tokens), making the representation biologically grounded and species-transferable. Trained on 36 million cells across diverse human tissues from the Tabula Sapiens consortium.
\n
Zero-shot cell type classification across tissues and species without fine-tuning. Embeddings capture biological relationships: similar cell types cluster together even across different organs and organisms. Published in Nature after extended peer review.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
The protein-embedding approach is what makes this work where others have not. When you represent genes as ESM embeddings of their protein products, you get biological similarity for free: genes with similar functions have similar representations regardless of what organism they come from. That is why UCE can transfer across species without retraining, while models using gene-name tokens cannot. The Nature publication after a 2023 preprint suggests the reviewers were convinced this is not just a benchmark artefact. For single-cell researchers, the practical question is whether UCE embeddings are useful enough to replace the fine-tuning workflows they already have with scGPT or Geneformer. The 33-layer model and code are public, so that comparison is straightforward to make.
The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.
ELIXIR\u2019s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals has now closed.
\n
\n
Thanks for reading!
\n
\ud83d\udcac Get involved
\n
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Connect With Us
\n
Have questions or suggestions? We'd love to hear from you!
Thanks for reading Kiin Bio! Subscribe for free to receive new posts and support my work.
\n
\n\n\n
\n
\n","enclosure":{"link":"https://substack-post-media.s3.amazonaws.com/public/images/62b01af7-9e9a-4b27-96fc-f6d5ea98a24e_1200x630.png","type":"image/jpeg"},"categories":[]},{"title":"Scigantic: The Platform That Democratizes Access to Large-Scale Data","pubDate":"2026-07-14 17:00:32","link":"https://newsletter.kiin.bio/p/scigantic-the-platform-democratizes","guid":"https://newsletter.kiin.bio/p/scigantic-the-platform-democratizes","author":"Natasha Kilroy","thumbnail":"","description":"Deep Dive | Edition 21","content":"\n
Welcome back to the deep dive, where we break down the AI tools and data reshaping how new drugs are discovered. In each edition, we speak directly with the teams behind these tools to explain what they solve, how they work and where they are going next.
\n
Open scientific data has an access problem, and a former Terra engineer is building the fix
\n
For computational biologists, genomics researchers, and any scientist working with large open-source datasets they can\u2019t actually use.
\n
Open data mandates have flooded repositories with petabytes of publicly available scientific data. In practice, most academic researchers can\u2019t touch it. Downloading costs thousands in egress fees, requires cloud infrastructure they don\u2019t have, and takes weeks of setup before a single analysis runs.
\n
Scigantic puts compute where the data already lives. Nearly 1 exabyte of open-source data appears as local files in a notebook, with zero transfer fees and built-in fine-tuning for foundation models like ESM and DNABERT. It was built by someone who spent years on Terra at the Broad Institute and saw exactly where enterprise genomics platforms fail the people who need them most.
\n
This piece covers what the platform does, why the person behind it is unusually well-positioned to build it, and what it means for the growing gap between \u201cdata exists\u201d and \u201cresearchers can work with it.\u201d
\n
\n
Kiin Pioneer Programme
\n
We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.
\n
The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
This week we spoke with Aaron Kanzer, founder of Scigantic, a platform that lets scientists work with massive open-source datasets without ever downloading them. Aaron is a solo founder running the company out of Boston, fully bootstrapped, with no outside funding. Before Scigantic he was a senior engineer at the Broad Institute working on Terra (the genomics platform serving 65,000 users across 80 petabytes of data), a founding engineer on MIT\u2019s LINC project mapping neural circuits, and most recently scaled ML operations at Generate:Biomedicines. He built Scigantic because he kept watching the same problem go unsolved from the inside.
The problem: open data that isn\u2019t actually open
\n
There is a quiet fiction in science right now. Funders mandate open data sharing. Repositories grow by petabytes each year. Papers cite datasets that are technically available to anyone. The word \u201copen\u201d appears in every grant report.
\n
The reality for a postdoc with a laptop and no cloud budget is different. \u201cOpen\u201d means the data exists on a server somewhere. To actually use it, you need to download terabytes over days or weeks, pay egress fees that can run into thousands of dollars per transfer, set up cloud infrastructure you were never trained to manage, and hope your local machine has the storage to hold it all. Most don\u2019t.
\n
The origin was personal. At MIT\u2019s McGovern Institute, Aaron was working on the LINC project, mapping neural circuits through high-resolution brain imaging. \u201cI noticed that these amazing datasets were being generated within the LINC project,\u201d he told us. \u201cIncredible high-resolution images of the brain with a \u2018wow\u2019 factor. However, the data engineering necessary to access these images was so technical and challenging.\u201d The data existed. It was extraordinary. And almost nobody outside a small group of engineers could actually get to it. \u201cHow might we improve this process for anyone in the scientific community?\u201d That question became Scigantic.
\n
The problem compounds with AI models. Tools like ESM and OpenFold are open-source, technically free to use. In practice, fine-tuning them on your own data requires weeks of infrastructure setup, or you go back to the original team and ask them to run it for you. The models are open. The ability to actually use them is not.
\n
As Kanzer puts it: \u201cOpen-source models are great as a base case for scientific discovery. If you have a more targeted hypothesis, or a specific therapeutic you are developing, fine-tuning becomes integral. Setting up fine-tuning though still has a steep data engineering learning curve before you can conduct any science.\u201d The bottleneck is not the model. It is everything between the model and actually using it on your own data.
\n
For context: AWS charges between $0.09 and $0.12 per gigabyte for data transfer out of S3. A single copy of the AlphaFold database is over 20 terabytes. That is roughly $2,000 just to download one dataset, before you\u2019ve done a single calculation. Multiply that across the dozens of datasets a real project might touch, and \u201copen\u201d starts to feel like a word with a hidden paywall attached.
\n
\n
The approach: compute moves to the data
\n
Scigantic inverts the model. Instead of moving data to the researcher, it moves the researcher\u2019s compute to where the data already lives. You open a JupyterHub notebook. The datasets you need appear as local files on your machine. You write code as if everything is sitting on your hard drive. Nothing ever downloads.
\n
The mechanism is elegant: FUSE mounts make cloud-hosted data appear as a local filesystem. Every file read translates to a range-GET against the object store, pulling only the bytes you actually need. Your analysis runs in the same cloud region as the data. When you\u2019re done, only the results leave: a trained model, a table, a figure. Typically megabytes, not terabytes.
A JupyterHub session on Scigantic. Open-source datasets appear as local directories in the file browser (left), while the built-in AI assistant (right) has context on every mounted dataset's schema.
\n
Kanzer\u2019s analogy is a library. \u201cRather than taking the entire book off the shelf, instead hand me the specific page I\u2019m looking for. No one is reading the entire book, so why move the entire book?\u201d That is what FUSE mounts achieve under the hood. The data stays in the cloud. Your notebook reads only the bytes it needs, as if they were sitting on a local drive. \u201cFUSE is great because it elegantly tricks the agent into operating as if the data is a local directory,\u201d Kanzer says. \u201cIt\u2019s a more clever mousetrap when dealing with large-scale datasets.\u201d
\n
For fine-tuning, the workflow is similarly stripped back. Upload a CSV of sequences and labels, pick a foundation model, train. LoRA on frozen backbones means the whole thing fits on 24-48 GB cards. No training loops to write. No dependency management. No fighting with CUDA versions at 2am.
The fine-tuning workflow: upload a CSV, pick a foundation model (here, ESMC-600M), and configure training. The platform validates your data format and warns about sample size before you start.
Completed fine-tuning run on ESMC-600M. Evaluation metrics, predicted vs. actual plots, and training curves are generated automatically. The trained model is available for download immediately.
Why it\u2019s different: built by the person who saw the limits from the inside
\n
What makes this more than another cloud notebook is who built it and why. Aaron spent time working on Terra at the Broad Institute, the largest open-source genomics platform in the world. He knows exactly what works about that system and exactly where it breaks down for the people who need it most. Terra is powerful, complex, enterprise-focused, and genomics-specific. If you\u2019re a neuroscientist or a climate researcher, it doesn\u2019t serve you. If you\u2019re a postdoc without a bioinformatics team behind you, the learning curve alone can take weeks.
\n
\u201cTerra is an amazing tool for researchers, but I felt it was tightly coupled to specific genomic hypotheses,\u201d Kanzer says. \u201cExploratory data analysis and fine-tuning require flexibility, so Scigantic is trying to abstract data engineering and ultimately let the researcher decide.\u201d The distinction matters. Terra serves genomics well. If you are a neuroscientist, a climate scientist, or simply someone who wants to explore data before committing to a hypothesis, it was not built for you. \u201cData-scaling issues are not isolated to genomics or life sciences,\u201d Kanzer adds. \u201cThey occur in all forms of science.\u201d
\n
One example of what \u201cdifferently\u201d looks like in practice: comparing ESM predictions against OpenFold outputs. Today, that means two separate environments, two sets of dependencies, two data pipelines. On Scigantic, it\u2019s one notebook session. Both models, same data, side by side.
ESM and OpenFold outputs compared in a single Scigantic notebook session. Two models, same data, no separate environments or dependency management required.
\n
The platform also includes an AI assistant built into JupyterLab that has context on every mounted dataset\u2019s schema. It can point you toward relevant columns, suggest analyses, or nudge you back if you\u2019re heading in an unproductive direction. It is not a chatbot bolted onto a product. It knows what data you\u2019re looking at.
\n
\n
Who it\u2019s for
\n
Scigantic is built for academics first. Postdocs, PhD students, researchers at institutions that don\u2019t have dedicated cloud engineering teams (which is most of them). The pricing reflects that: a free tier exists, and the researcher plan is $19 per month.
\n
The logic is deliberate, not a limitation. \u201cThe most impactful scientific breakthroughs originate in academia before being commercialised,\u201d Kanzer says. \u201cThe Human Genome Project. The Protein Data Bank. Scigantic believes that prioritising academics first will lead to a much larger impact later on.\u201d Today\u2019s postdoc running analyses on the free tier is tomorrow\u2019s head of computational biology choosing infrastructure for an entire department.
\n
Early feedback suggests the platform points researchers in productive directions faster than existing tools. The AI assistant in particular seems to help people who know what question they want to ask, but don\u2019t know which dataset or which column holds the answer.
\n
Early signs suggest it is working. One MIT postdoc put it directly: \u201cScigantic\u2019s ability to quickly navigate large S3 buckets is incredibly valuable. Gemini, Claude, Codex, etc. try to download large datasets in real-time, often never getting a solid answer to my question. I\u2019m thoroughly impressed how fast it gets me onboarded to the data.\u201d The bar here is not perfection. It is whether the platform gets a researcher to a productive starting point faster than the alternatives. By that measure, it seems to be landing.
\n
\n
The future
\n
Aaron has started conversations with multiple renown research institutions. He\u2019s exploring what a partnership with Kiin could look like. The vision is to become the default access layer between open scientific data and the researchers who need it, across every domain, not just genomics.
\n
\u201cWe see multiple partnerships forming with Scigantic in the next year, further validating our hypothesis that navigating unstructured, large-scale data is the bottleneck,\u201d Kanzer says. NASA conversations are already underway. The thesis is simple: if the bottleneck is access, and access keeps getting harder as datasets grow, then the platform that solves it once becomes the default layer everyone builds on.
\n
The fact that this is bootstrapped and solo matters. There\u2019s no board pushing toward enterprise sales or premature monetisation. The roadmap can stay focused on what academics actually need, for now. That is a strategic choice, not a limitation.
The \u201cwhy now\u201d here is straightforward. Open data mandates from NIH, Wellcome, and the EU are flooding repositories with petabytes of new data every year. Foundation models for biology are proliferating faster than any lab can keep up. The gap between \u201cdata exists\u201d and \u201cI can work with it\u201d is widening, not closing. Someone was going to build this access layer. The fact that it\u2019s being built by someone who spent years inside Terra, who understands both the infrastructure and the user pain at a level most founders don\u2019t, is what makes it credible.
\n
The risk is execution at scale with a single person. The opportunity is that the product is simple enough in concept (notebook + mounted data + fine-tuning) that it doesn\u2019t need a 50-person engineering team to work. It needs to work reliably for the 10,000 postdocs who currently can\u2019t access the data their own field generated. If it does that, the enterprise customers will follow.
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.
\n
Connect With Us
\n
Have questions on this or suggestions for our next deep dive? We\u2019d love to hear from you!
Welcome back to your weekly dose of AI news for Life Science! This weeks fix:
\n
\n
C3P uses proteins as a training signal to learn promoter representations, which lets you find co-regulated genes across bacterial genomes without any experimental data. Simple idea, works well.
\n
BioMatrix is another biological foundation model, but this one actually covers the full matrix: molecule sequences, molecule structures, protein sequences, protein structures, and natural language, all in one model. State-of-the-art or competitive on 77 of 80 benchmarks.
\n
JEDEL automates DNA-encoded library design from a pharmacophore, and everything it generates is synthesisable from purchasable building blocks. That constraint is what makes it useful rather than academic.
\n
\n
\n
Kiin Pioneer Programme
\n
We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.
\n
The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
Genome language models (like DNABERT, Nucleotide Transformer, Evo) have gotten good at learning sequence representations, but they still struggle with regulatory DNA. Promoters are short, non-coding, and their function depends on context that pure sequence models have trouble capturing. The standard approach is to pretrain on raw DNA and hope the model picks up regulatory grammar along the way. It mostly does not.
\n
C3P takes a different angle entirely. Rather than trying to learn promoter function from DNA sequence alone, it uses the protein that a promoter regulates as a supervisory signal. The logic: promoters that regulate similar functions should have similar representations, and protein language models already capture functional similarity well. So you can transfer that knowledge to the promoter side through contrastive learning. It is a simple idea, and arguably obvious in hindsight, but nobody has done it at this scale before.
Genome language models learn poor representations of regulatory DNA, limiting their use for predicting gene regulation. Dufault, Xu, and Moses from the University of Toronto present C3P, a contrastive learning framework that pairs bacterial promoters with their downstream proteins to learn promoter representations.
\n
Trained on 88 million promoter-protein pairs using a CLIP-style objective. The promoter encoder learns to align with embeddings from a frozen protein language model, so promoters regulating functionally similar proteins end up with similar representations.
\n
Multi-fold improvement over leading genome language models on regulatory annotation prediction. Enables zero-shot co-regulated gene retrieval: given a promoter, find other promoters driving similar functions, across genomes, with no experimental data required.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
The clever bit here is recognising that we already have a strong signal for regulatory function, it just lives in protein space rather than DNA space. Protein language models have encoded functional relationships that took decades of biochemistry to establish. C3P bridges that knowledge back to the regulatory side. This matters for microbiology because most bacterial genomes have no experimental regulatory data at all. If this approach generalises beyond the training distribution (which the zero-shot retrieval results suggest it might), it opens up regulatory annotation for millions of uncharacterised organisms. The limitation is that it only works where you have a clear promoter-protein pair, which excludes non-coding RNAs and complex eukaryotic regulation.
The biological foundation model space has been fragmented. You have protein language models (ESM, ProtTrans), molecule models (MolBERT, ChemBERTa), structure predictors (AlphaFold, ESMFold), and various attempts to combine two of these modalities. What nobody has done convincingly is put all five modalities (molecule sequence, molecule structure, protein sequence, protein structure, and natural language) into one architecture that can both read and generate all of them. Previous multi-modal attempts (BioMedGPT, Galactica) either used modality-specific encoders bolted onto a language model, or covered text plus one other modality.
\n
BioMatrix from Tsinghua and collaborators argues you do not need specialised encoders at all. You tokenise everything into a shared vocabulary, train with standard next-token prediction, and let the model sort out the relationships. The scale helps: 304 billion tokens across all modalities, built on Qwen3 at 1.7B and 4B parameters.
Existing biological AI models are specialised to one or two data types and cannot transfer knowledge between molecules and proteins natively. Pei et al. from Tsinghua present BioMatrix, a multimodal foundation model that unifies molecule sequences, molecule structures, protein sequences, protein structures, and natural language in a single architecture.
\n
All modalities are mapped to a shared token space using learned structure tokenisers (VQ-VAE for molecular and protein 3D structures). The model is then trained with a standard next-token prediction objective, with no external encoders or modality-specific output heads. Built on Qwen3, trained on 304.4 billion tokens.
\n
State-of-the-art or competitive on 77 out of 80 downstream tasks across property prediction, molecule generation, protein folding, captioning, and binding affinity prediction. Outperforms modality-specific specialist models on several benchmarks.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
The interesting question is not \u201cdoes a bigger model do well on benchmarks\u201d (it does, unsurprisingly) but whether unifying modalities in one model creates emergent cross-modal capabilities that specialist models cannot replicate. The paper does not fully answer this yet, and most of the 80 tasks are single-modality evaluations that a specialist could handle. The real test will be tasks that require reasoning across modalities simultaneously: \u201cgiven this protein structure and this molecule, predict binding and explain why.\u201d If BioMatrix can do that better than a pipeline of specialists, the unified approach is vindicated. If it just matches specialists on their own turf, the value proposition is convenience rather than capability. The Apache 2.0 license and public weights make it easy to test either way.
DNA-encoded libraries (DELs) are combinatorial chemistry at industrial scale: you attach DNA barcodes to building blocks, combine them through validated reactions, and screen the resulting millions of compounds against a target. The design problem is choosing which building blocks and reactions to include. Traditional DEL design relies on chemical diversity heuristics or brute-force enumeration, neither of which accounts for whether the resulting library will actually produce binders for your specific target.
\n
Generative drug design models (like those from Recursion, Insilico, etc.) can propose target-specific molecules, but they produce virtual structures that then require separate synthesis planning, which often fails. JEDEL closes this gap by designing libraries that are target-aware and synthesisable by construction: every output is built from purchasable building blocks through validated reactions.
Current DEL design ignores target information, and generative models propose molecules that are often unsynthesisable. Jocys et al. from the University of Surrey present JEDEL, a framework that converts 3D pharmacophore information from known active ligands into synthesis instructions for target-focused DNA-encoded libraries.
\n
Takes pharmacophore representations as input, searches a space of purchasable building blocks and validated combinatorial reactions, and assembles libraries that are synthesisable by definition. Requires no target-specific retraining, operating in a zero-shot manner across different protein targets.
\n
Tested across 18 protein targets, JEDEL outperformed random and diversity-based baselines on predicted binding affinity, pharmacophore recovery, and sample efficiency. The constraint to purchasable reagents means every designed library can be made in the lab immediately.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
DELs are already a workhorse in pharma (Novartis, GSK, and X-Chem all run them at scale), so improvements here have a clear commercial path. What JEDEL does is shift library design from \u201cmaximise chemical diversity and hope something binds\u201d to \u201cbuild a library biased toward your target\u2019s pharmacophore.\u201d The zero-shot aspect means you do not need to retrain for each new campaign. The synthesisability constraint is the real differentiator from generative models: the gap between \u201chere is a molecule that might bind\u201d and \u201chere is a library you can make on Monday\u201d is where most computational drug design papers lose their translational value. The limitation is that pharmacophore inputs require existing active compounds, so this is a hit expansion tool rather than a de novo discovery method. For teams already running DEL campaigns, this looks immediately applicable.
No public code repository available at time of writing.
\n
\n
\ud83d\uddd3\ufe0f Events & Competitions
\n
The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.
ELIXIR\u2019s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals has now closed.
\n
\n
Thanks for reading!
\n
\ud83d\udcac Get involved
\n
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Connect With Us
\n
Have questions or suggestions? We'd love to hear from you!
Thanks for reading Kiin Bio! Subscribe for free to receive new posts and support my work.
\n
\n\n\n
\n
\n","enclosure":{"link":"https://substack-post-media.s3.amazonaws.com/public/images/4106577f-f94a-46f7-82c1-781907961b71_1200x630.png","type":"image/jpeg"},"categories":[]},{"title":"A Primer on Molecular Docking","pubDate":"2026-06-23 17:01:56","link":"https://newsletter.kiin.bio/p/a-primer-on-molecular-docking","guid":"https://newsletter.kiin.bio/p/a-primer-on-molecular-docking","author":"Natasha Kilroy","thumbnail":"","description":"Why docking scores don\u2019t predict binding affinity, and what template-guided approaches actually deliver","content":"\n
Welcome back to Kiin Bio Weekly.
\n
\n
For computational chemists, drug discovery scientists, and anyone running or interpreting docking campaigns.
\n
Docking scores do not predict binding affinity. The OpenBind benchmark confirmed it: molecular weight alone outperformed Gnina, Boltz-2, and other state-of-the-art models on experimental binding data.
\n
The real value of docking is pose prediction, not ranking compounds. Most teams using it well have stopped asking it to estimate affinity altogether.
\n
This piece covers where docking fails, where it works (template-guided lead optimisation), and why pooling imperfect methods beats waiting for a perfect one.
\n
\n
\n
Freebie alert: We know how hard science is. That\u2019s why we built the Pioneer Programme.
\n
We\u2019re selecting academic and nonprofit teams to get one year of free access to our drug discovery platform, with support from our science team. If you spend more time pulling together findings from different sources than actually acting on them, it\u2019s worth applying.
\n
No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
For this piece, I spoke with Rachael Skyner, our cheminformatician at Kiin Bio, about the mechanics of molecular docking, why scoring remains unsolved, and how template-guided approaches are changing what is possible in lead optimisation. Her insight runs throughout.
\n
\u201cThe search problem is not the hard part. It\u2019s ranking them and picking the right pose.\u201d
\n
\n
Why this matters now
\n
Thanks to AlphaFold and its successors, drug discovery teams now have predicted structures for essentially every human protein. Generative chemistry tools are producing novel molecules faster than medicinal chemists can evaluate them. Having a structure to dock into is no longer the problem. Trusting what the docking tells you is. The OpenBind benchmark, published earlier this year, put hard numbers on this trust deficit for the first time, and the results were not encouraging.
\n
Docking can tell you roughly where a molecule might land. It cannot reliably tell you how tightly it will stick. Most practitioners have adapted by splitting the problem in two: get the pose right first, worry about affinity later. That distinction shapes everything about how docking is actually used today.
Figure 1. The molecular docking workflow: ligand and receptor are prepared, the ligand is docked into the binding site across multiple orientations, poses are scored, and the best pose is selected. The scoring and selection steps are where most methods struggle.
\n
\n
The scoring problem nobody has solved
\n
Traditional docking scores are supposed to estimate binding affinity. They do not. The physics approximations used in scoring are rough even for a static system, and on top of that, docking ignores the fact that binding is actually a dynamic process. Binding in the body involves solvent reorganisation, protein conformational changes, entropy costs, and timescale-dependent interactions that no static pose can capture.
\n
\u201cIt\u2019s an oversimplification of what\u2019s actually going on with binding,\u201d says Skyner. \u201cIt\u2019s not just binding that contributes to affinity in the end. There are lots of other things that can influence how strongly a molecule binds. You\u2019re treating it as a static problem when in reality it\u2019s very dynamic.\u201d
\n
Increasingly, people are separating pose prediction from affinity estimation altogether. Modern scoring methods from tools like Gnina use convolutional neural networks trained on the PDBbind dataset. Rather than predicting binding affinity directly, the CNN score predicts confidence in whether the pose is correct, which is a more tractable and more useful question to ask.
Figure 2. Binding affinity prediction from the OpenBind benchmark. Molecular weight (a trivial baseline) correlates better with experimental affinity than scores from Gnina, Boltz-2, and other structure-based methods.
\n
\n
Protein flexibility and the induced fit problem
\n
Traditional rigid docking assumes the protein does not move, which is obviously wrong. Proteins reshape themselves around ligands in a process called induced fit, and this is why rigid receptor docking often fails for novel chemotypes.
\n
Physics-based approaches like Rosetta and Schrodinger\u2019s IFDMD allow limited residue rotation or flipping at the binding site. Machine learning methods like DiffDock take a generative approach, producing the molecule directly into the protein. Co-folding methods fold the protein and ligand simultaneously, addressing flexibility implicitly by allowing both to move at the same time during prediction.
\n
The problem is that induced fit methods produce multiple protein conformations alongside multiple ligand conformations, and the analysis becomes exponentially more complex. \u201cFrom an analysis point of view, it\u2019s really difficult to deal with those structures,\u201d says Skyner. \u201cYou have to start doing more complicated things like clustering together both the ligand and the protein conformation to see which ones come up most often.\u201d
\n
For high-throughput virtual screening where you want to process hundreds of thousands of molecules, this complexity is impractical. Induced fit docking belongs in the later stages of a project, when you already have a binder and want hypotheses about its binding mode.
\n
Traditional rigid docking assumes the protein does not move. Proteins move constantly. They reshape themselves around ligands in a process called induced fit, and this is why rigid receptor docking often fails.
\n
\n
The virtual screening funnel
\n
In practice, docking is rarely used alone. It sits inside a virtual screening funnel: start with a compound library (thousands to millions of molecules), run fast docking as a heuristic filter (anything scoring worse than minus eight kilocalories per mole gets discarded), then rescore the surviving hits with more expensive methods.
\n
Rescoring typically involves energy minimisation with solvent. Methods like MMGBSA and MMPBSA incorporate water into the calculation, which basic docking ignores entirely. \u201cWhen you start thinking about the solvent as well, you start getting much better correlation with experimental affinities,\u201d says Skyner. These methods are computationally expensive, though still feasible for the few hundred molecules that survive initial filtering.
\n
ML-based rescoring adds another option: build a local model of the docking score and then only dock the subset your model flags as likely to score well. This can cut the number of molecules you need to physically dock in half.
\n
\n
Where docking actually works: template-guided lead optimisation
\n
Docking works best in template-guided approaches during lead optimisation, not in blind global searches or massive virtual screens. This is where Skyner\u2019s work focuses.
\n
At this stage, you already have a molecule with known binding evidence. You are working through a congeneric series: molecules that share a common core (the maximum common substructure, or MCS) with small modifications around the periphery. A chlorine added here, a fluorine swapped there, an atom changed in a ring. You already know the molecule binds. What you want to understand is whether the modification changes how it sits in the pocket.
\n
Template-guided docking uses the known crystal structure of a reference compound to anchor the search. Rather than exploring the entire binding pocket from scratch, you start from the position of the known binder and run a local minimisation. Skyner has developed an MCS-guided docking approach that generates low-energy conformers of the ligand, rigidly aligns them to the reference structure via the MCS, and performs local optimisation from that aligned position using AutoDock Vina.
\n
\u201cIf you give it the right starting position, it\u2019s got a much better chance of getting the pose correct,\u201d says Skyner. \u201cPeople might do this by default, but it really, really helps in these scenarios.\u201d
Figure 3. Global search vs template-guided docking. In global search, the algorithm explores the entire binding pocket from random starting positions. In template-guided docking, the known crystal structure of a similar molecule anchors the search, with the congeneric series sharing a maximum common substructure (MCS) highlighted.
\n
\n
Ensembles, pooling, and what comes next
\n
No single docking method dominates across all targets. In Skyner\u2019s evaluation of 22 targets (1,888 structures), pooling results from multiple methods and rescoring with Gnina\u2019s CNN score consistently outperformed any individual approach. The pooled \u201coracle\u201d (checking whether any method generated the correct pose) found it 90% of the time using RMSD <2 angstroms as the heuristic for a correct binding pose, and 54% by the more stringent SuCOS metric, which evaluates shape and pharmacophore overlap rather than just atomic positions.
\n
The practical recommendation: combine Vina (fast, global search) with MCS-guided Vina (template-aware), pool the poses, and rescore with the CNN score. Two methods captured nearly all the benefit of three, with substantially less compute.
Figure 4. Product scoring combines CNN confidence (y-axis) with SuCOS shape/pharmacophore overlap (x-axis) to select the best poses. Poses in the upper-right quadrant score highly on both criteria and are most likely to be correct. Different targets show different distributions, reflecting real binding mode diversity across series.
\n
\n
The preparation of proteins and ligands is still the most error-prone step in any docking workflow. Get the protonation states wrong, miss a stereoisomer, assign incorrect charges, and the docking algorithm has no chance of finding the right answer. Errors compound downstream into molecular dynamics and other physics-based follow-up methods. \u201cIf you get it wrong from the beginning, it really affects everything else,\u201d says Skyner.
\n
Docking will not predict your binding affinities, and it will not reliably rank compounds within a series. The industry spent years asking it to do both, and the benchmarks now confirm what practitioners quietly knew. The gains are coming from asking docking the right question in the right context: use it for pose prediction rather than affinity estimation, use template-guided exploration rather than blind search, pool results from multiple methods rather than betting on one. The teams actually getting value out of docking today have mostly just stopped asking it to be something it was never designed to be.
\n
\n
\ud83d\udcac Want to be featured in Kiin Bio Weekly?
\n
Each issue we speak directly with researchers, scientists, and builders working at the frontier of AI in life sciences. If you're working on something in this space and think it would resonate with our community, I'd love to hear from you. Fill out this form or reach out to me directly.
\n
\n
Found this useful? Forward it to a colleague in computational biology, it's the best way to help the newsletter grow.
Subscribe now to stay at the forefront of AI in Life Science. Every week: primers, deep dives, and direct conversations with the people building the field.
Welcome back to your weekly dose of AI news for Life Science! This weeks fix:
\n
\n
Google\u2019s AMIE system matched or beat primary care physicians on clinical management reasoning across 100 multi-visit scenarios. Published in Nature, and the study design is more rigorous than most in this space.
\n
CANVAS turns standard H&E slides into virtual spatial proteomics maps, predicting tumour microenvironment neighbourhoods from cheap histology. Validated across 5,000 patients and 9 cancer types.
\n
AMPGAN v3 is the first generative model for antimicrobial peptides that handles non-canonical amino acids and chemical modifications. Two of five generated candidates showed real antimicrobial activity. They also built an agentic pipeline around it, which is where this gets interesting.
\n
\n
\n
Kiin Pioneer Programme
\n
We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.
\n
The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
Most AI-in-medicine papers test whether a model can diagnose from a static vignette. That is a solved problem at this point, or at least a well-explored one. What has not been tested seriously is whether an AI system can manage a patient over time: adjust treatment plans across multiple visits, respond to new lab results, and prescribe medications safely. That is what primary care actually involves, and it is where AMIE (Articulate Medical Intelligence Explorer) from Google DeepMind now enters.
\n
Previous AMIE work showed the system could match physicians on diagnostic conversations. This paper extends it to management reasoning, which is harder because it involves sequential decisions, guideline interpretation, and medication safety. The comparison set is interesting: 21 board-certified primary care physicians across 100 multi-visit case scenarios grounded in UK NICE and BMJ Best Practice guidelines.
Management reasoning for longitudinal care is largely untested for AI. Li\u00e9vin et al. from Google DeepMind present an agentic version of AMIE optimised for multi-visit clinical management and medication reasoning.
\n
Uses Gemini\u2019s long-context window to combine in-context retrieval of clinical guidelines with structured reasoning. A multi-agent architecture handles dialogue, management planning, and medication lookup from drug formularies (OpenFDA, BNF).
\n
In a blinded virtual OSCE study, AMIE scored significantly higher than PCPs on treatment preciseness (96% vs. 62%, p<0.001) and guideline alignment (93% vs. 75% by visit 3). On the RxQA medication benchmark, AMIE outperformed PCPs on harder questions (57.9% vs. 47.8%, p<0.001). Specialist physicians and patient actors preferred AMIE 47% of the time vs. 7% for PCPs.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
The honest reaction here is: this is impressive and somewhat uncomfortable. AMIE is not just pattern-matching against guidelines, it is reasoning about how to adjust a plan given what happened at the last visit. The study design (blinded OSCE, specialist evaluators, real guidelines) is more credible than most in this space. The medication reasoning results are particularly notable because prescribing errors are a leading cause of preventable harm. That said, this is still a virtual scenario, not a real clinic with real patients who do unexpected things. The gap between \u201cperforms well in structured evaluation\u201d and \u201ccan safely manage my mum\u2019s hypertension\u201d remains large. What this does establish is that the technical capability exists. The regulatory and deployment questions are now the binding constraint, not the model performance.
Spatial proteomics (CODEX, MIBI, etc.) can map the tumour microenvironment at single-cell resolution, but it costs thousands per sample and requires specialised equipment. Standard H&E histopathology costs almost nothing and is already collected for every cancer patient. The question is whether you can infer the spatial biology from the cheap stain. Previous attempts have tried to predict individual protein expression from H&E, but that approach is fragile: sensitive to staining variation, limited to a handful of markers, and accuracy drops quickly. CANVAS takes a different approach, predicting cellular neighbourhood patterns rather than individual proteins, which is a more robust prediction target.
\n
This connects to a broader trend in computational pathology where foundation models (UNI, Virchow, CONCH) have made feature extraction from H&E much more powerful, and the question is now what downstream tasks those features can support. CANVAS uses these pretrained features to bridge modalities rather than training from scratch.
Spatial proteomics reveals the tumour microenvironment in detail but costs too much to scale. Li et al. from Stanford present CANVAS, an AI platform that predicts spatial cellular neighbourhood structures directly from standard H&E slides, trained on an atlas of 18 million cells profiled by 41-plex CODEX imaging.
\n
CANVAS defines 10 reproducible cellular neighbourhoods from CODEX data across 457 lung cancer patients, then trains a pathology foundation model to predict these neighbourhood patterns from co-registered H&E images. It operates at the ecological niche level rather than individual cell types.
\n
Applied to over 5,000 patients across 9 cancer types, CANVAS-derived spatial features predicted immunotherapy response with AUCs above 0.75 at 6, 12, and 24 months. The spatial signature stratified patients by progression-free survival (HR = 2.42, p<0.001) and outperformed established biomarkers including TMB, PD-L1 expression, and TLS. Validated externally on a Cancer Moonshot Biobank cohort.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
This matters for a specific reason: immunotherapy response prediction is a clinical problem where existing biomarkers (PD-L1, TMB) work poorly. About 20-30% of patients respond to checkpoint inhibitors, and we are bad at predicting who they will be beforehand. CANVAS proposes that the spatial organisation of the tumour microenvironment, inferred from a slide that already exists in every pathology lab, is more informative than the molecular markers we have been relying on. If the external validation holds up across broader cohorts (the Moonshot cohort is small at n=40), this could actually change who gets prescribed immunotherapy. The non-commercial license limits immediate industry adoption, but for academic cancer centres this is usable now.
Antimicrobial resistance causes over a million deaths annually, and no new antibiotic class has been commercialised since 2000. Antimicrobial peptides (AMPs) are attractive because they disrupt membranes through physical interactions, making resistance harder to develop. Generative models for AMP design exist (PepGAN, HydrAMP, AMP-Designer), but they all share two limitations: they only work with natural L-amino acids, and they require manual filtering of outputs. Real therapeutic peptides need D-amino acids and terminal modifications to survive in the body. AMPGAN v3 is the first generative model that handles these non-canonical chemistries, and PepCraft wraps it in a multi-agent pipeline that automates the filtering.
\n
The field has been generating lots of candidate peptides computationally, but the translation gap to actual antimicrobials has been wide. Most papers stop at predicted activity scores. This one synthesises candidates and tests them against real bacteria, which is the bar that matters.
Generative models for antimicrobial peptides are limited to natural amino acids and produce outputs that need extensive manual curation. Jung et al. from the University of Vermont and Purdue present AMPGAN v3, a conditional GAN that generates antimicrobial peptides with D-amino acids and terminal modifications, paired with PepCraft, a multi-agent framework for automated AMP discovery.
\n
AMPGAN v3 separates adversarial training across two discriminators: one for sequence realism, one for antimicrobial activity prediction. This fixes the training instability that plagued earlier versions (only ~10% of AMPGAN v2 runs produced usable models). PepCraft uses a Planning Agent to coordinate specialised executors for generation, physicochemical filtering, and database verification.
\n
Two of five synthesised candidates showed clear antimicrobial activity against Gram-positive strains, with the best reaching MIC of 8 \u03bcg/mL against B. subtilis. The candidates spanned three structural classes (alpha-helical, beta-hairpin, random coil) and incorporated D-amino acids and amidation, which previous generative methods cannot produce. PepCraft\u2019s prioritisation recommendations aligned with the wet-lab results.
\n
\n
\ud83d\udca1 Why This Is Cool
\n
The wet-lab hit rate (2/5) is respectable for a generative model, and the chemical space expansion is the real contribution. Every other AMP generator is restricted to natural amino acids, which means their outputs degrade rapidly in serum. Expanding the vocabulary to include D-amino acids and terminal caps makes the generated peptides actually viable as therapeutics rather than just interesting sequences. The agentic pipeline (PepCraft) is early-stage and exploratory, but it points toward a pattern we will see more of: generative models wrapped in verification agents that can filter, cross-reference, and prioritise without human intervention. This was accepted at the ICML 2026 GenBio workshop, not a top venue, and the validation is limited to Gram-positive bacteria. Worth watching, not yet proven at scale.
The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.
A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. \u00a360 for biotech companies.
The first edition of a one-day symposium bringing together London\u2019s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub L\u00e1la. In person only.
ELIXIR\u2019s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals has now closed.
\n
\n
Thanks for reading!
\n
\ud83d\udcac Get involved
\n
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Connect With Us
\n
Have questions or suggestions? We'd love to hear from you!
Thanks for reading Kiin Bio! Subscribe for free to receive new posts and support my work.
\n
\n\n\n
\n
\n","enclosure":{"link":"https://substack-post-media.s3.amazonaws.com/public/images/47699992-396f-423e-b55e-6d8a521c4c49_1200x630.png","type":"image/jpeg"},"categories":[]},{"title":"UPenn's mRNAutilus, GSU's EpiFormer, and Seoul National's Folddisco","pubDate":"2026-06-11 17:02:08","link":"https://newsletter.kiin.bio/p/dukes-mrnautilus-asus-epiformer-and","guid":"https://newsletter.kiin.bio/p/dukes-mrnautilus-asus-epiformer-and","author":"Natasha Kilroy","thumbnail":"","description":"Kiin Bio's Weekly Insights","content":"\n
Welcome back to your weekly dose of AI news for Life Science! This weeks fix:
\n
\n
mRNAutilus generates entire therapeutic mRNA sequences from scratch, and the wet-lab numbers are hard to argue with: 400x over wild-type expression, beating commercial Spike constructs.
\n
EpiFormer brings geometric deep learning to epitope prediction with a 40% F1 boost. A nice complement to last week\u2019s ESM binder design coverage, now from the antigen side.
\n
Folddisco indexes 53 million protein structures and finds structural motifs in seconds. The Steinegger lab keeps quietly building infrastructure that makes everyone else\u2019s work faster.
\n
\n
\n
Kiin Pioneer Programme
\n
We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.
\n
The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
\ud83d\udd2c Designing full-length therapeutic mRNAs means optimising stability, translation efficiency, and codon usage simultaneously. Current methods tackle these objectives piecemeal, stitching together separately optimised UTRs and coding regions, which leaves performance on the table.
\n
Patel et al. from the Chatterjee lab at Duke present mRNAutilus, a generative framework that designs complete mRNA transcripts optimised across multiple properties at once.
\n
\ud83e\uddec The system trains a masked discrete diffusion model on millions of full-length mRNAs, then steers generation with Monte Carlo tree guidance to hit multiple objectives without retraining. It operates on whole transcripts rather than modular components.
\n
\u26a1 Zero-shot mRNAutilus designs encoding firefly luciferase achieved over 400-fold higher expression than wild-type, outperforming commercial baselines. For SARS-CoV-2 Spike, designs matched or surpassed both clinically used constructs and lab-optimised sequences. The framework also generalised to prime editing guides and targeted protein degradation, which suggests this is not a one-trick benchmark result.
mRNA therapeutics have had a sequencing problem disguised as a design problem. Post-COVID, the bottleneck is no longer \u201ccan we make mRNA drugs?\u201d but \u201ccan we make them well enough to compete on potency and manufacturing cost?\u201d Tools like LinearDesign (Zhang lab, 2023) optimise coding sequences for stability, and UTR-focused approaches pick regulatory elements, but these treat the transcript as a set of independent modules. mRNAutilus is the first framework I have seen that treats the entire transcript as a single generative object, which matters because interactions between UTRs and coding regions affect folding and translation in ways modular approaches miss.
\n
The wet-lab validation is what separates this from yet another generative model paper. Beating commercial constructs for Spike expression is a meaningful bar, not an in silico benchmark. The generalisation to prime editing and degraders suggests the architecture is flexible enough to not be overfit to reporter assays. The timing makes sense too: masked diffusion models have matured enough (thanks to protein and genomics applications) that applying them to mRNA sequences is a natural next step, and the Monte Carlo tree guidance borrows from AlphaGo-era decision strategies to handle multi-objective trade-offs without expensive retraining.
\n
For readers working in mRNA therapeutics: this is worth watching closely. The code is not yet public, which limits immediate adoption, but the approach could compress the design-test cycle considerably once available.
\n
\ud83d\udca1 Why This Is Cool
\n
The shift from \u201coptimise one property\u201d to \u201cgenerate the whole thing optimised\u201d matters more than it sounds. The history of biologics design is littered with tools that optimised one metric while inadvertently breaking another. If multi-objective generation holds up across more constructs and delivery contexts, it moves mRNA design closer to what protein design achieved with diffusion models over the past two years. The open question is whether the approach scales to longer, more complex transcripts and novel target classes beyond the well-studied ones shown here.
\ud83d\udd2c Predicting which surface residues an antibody will target on an antigen remains stubbornly difficult. Most existing methods treat the antigen in isolation, ignoring the antibody entirely or bolting antibody information on as a late-stage afterthought.
\n
Ahmed et al. from Georgia State University introduce EpiFormer, a geometric deep learning framework that models antigen-antibody interactions through interleaved cross-attention within GNN encoding layers.
\n
\ud83e\uddec Rather than encoding antigen and antibody separately then combining representations at the end, EpiFormer threads cross-attention between the two structures at every encoding layer. This allows bidirectional information flow throughout the representation, so the model learns how antibody geometry constrains which epitope residues are accessible.
\n
\u26a1 On standard benchmarks, EpiFormer achieves over 40% improvement in F1 score compared to previous best methods. That is a substantial jump for a prediction task where incremental gains of 2-5% have been typical. The model operates on 3D structural inputs from antibody-antigen complexes.
Epitope prediction sits upstream of antibody engineering: if you know where an antibody binds, you can design better binders and prioritise vaccine targets. Previous approaches like DiscoTope and ElliPro use geometry and surface properties of the antigen alone, which is a bit like predicting where a key fits without looking at the lock. More recent methods (PECAN, AsEP) incorporated paratope information but typically as a separate encoding step with late fusion.
\n
EpiFormer\u2019s contribution is architectural rather than data-driven. The interleaved cross-attention ensures antibody context informs antigen representations from the start rather than being concatenated at the decision layer. This connects naturally to last week\u2019s coverage of ESM-based binder design: that work generates antibodies given a target, while EpiFormer predicts where on the target those antibodies will land. Together they cover both directions of the same binding prediction problem.
\n
The 40% F1 improvement is striking, though it warrants some caution. Epitope prediction benchmarks are notoriously sensitive to train/test splitting, and structural epitope datasets remain small (a few thousand complexes in SAbDab). Whether this holds on truly novel antigen folds or just reflects better exploitation of known structural patterns is an open question. The code is available, which helps.
\n
\ud83d\udca1 Why This Is Cool
\n
The \u201cinterleave information early rather than fuse late\u201d lesson keeps appearing across structural biology. AlphaFold did it for MSA and structure tracks. ESM3 does it for sequence, structure, and function. EpiFormer applies the same intuition to a paired prediction problem. The field has been underestimating how much cross-modal information gets lost in late-fusion architectures, and each new result in this direction makes that clearer. For antibody discovery teams, this is immediately useful if it generalises beyond the benchmark setting.
\ud83d\udd2c Finding recurring 3D structural motifs (zinc fingers, catalytic triads, protein-protein interaction surfaces) across millions of predicted structures is computationally prohibitive. Existing methods either cannot handle discontinuous motifs or choke on databases beyond a few hundred thousand structures.
\n
Kim et al. from the Steinegger lab at Seoul National University present Folddisco, a structural motif search tool that indexes 53 million AFDB50 structures in a 1.45 TB index and returns query results in seconds.
\n
\ud83e\uddec Folddisco encodes proximal residue pairs into geometric feature sets (distances, angles, and side-chain orientation via torsion angles), stores them in a position-independent inverted index, and ranks hits using an IDF-based coverage score that rewards rare features. This handles both short continuous motifs and long discontinuous ones.
\n
\u26a1 Indexing is 11x faster to build and 4x more storage-efficient than previous state-of-the-art. Query speed is 20-fold faster than pyScoMotif on the full pipeline. On the zinc finger benchmark against the human proteome, Folddisco outperformed both RCSB and pyScoMotif on recall while maintaining higher precision. It also successfully distinguished active from inactive GPCR conformational states using activation motifs.
This is infrastructure work, and the most consequential kind. The AlphaFold database gave us 200+ million predicted structures, but searching them structurally has lagged far behind searching them by sequence (where tools like Foldseek, also from the Steinegger lab, already operate at scale). Folddisco fills the motif-search gap: given a 3D pattern of interest, find everywhere it occurs across all known and predicted protein structures.
\n
The practical value becomes clear with the GPCR example. Being able to query \u201cshow me all structures with this activation motif\u201d across both experimental PDB structures and AlphaFold predictions means you can study conformational states at proteome scale. That was previously manual curation work. Similarly, the zinc finger detection in uncharacterised metagenomic proteins (from ESM30) demonstrates functional annotation where sequence-based methods fail entirely.
\n
Folddisco\u2019s main limitation is its 20-angstrom connectivity constraint, which means it cannot detect spatially distant functional sites like remote allosteric pockets. The IDF scoring also struggles with very short motifs. These are known trade-offs for the speed gains.
\n
\ud83d\udca1 Why This Is Cool
\n
The Steinegger lab has been building the search infrastructure for the structure-prediction era piece by piece: MMseqs2 for sequences, Foldseek for structure alignment, and now Folddisco for motif search. Each tool makes the previous one more useful. What matters here is not the individual benchmarks but the fact that motif search at 53-million-structure scale is now a webserver query rather than a compute cluster job. That means anyone with a structural intuition and a browser can generate hypotheses that previously required a compute cluster and custom code. The webserver is live at search.foldseek.com/folddisco.
\n
\ud83d\udcc3 Read the paper. \ud83d\udcbb Try the code.
\n
\n
\ud83d\uddd3\ufe0f Events & Competitions
\n
The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.
A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. \u00a360 for biotech companies.
The first edition of a one-day symposium bringing together London\u2019s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub L\u00e1la. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.
ELIXIR\u2019s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that\u2019s your window.
\n
\n
Thanks for reading!
\n
\ud83d\udcac Get involved
\n
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Connect With Us
\n
Have questions or suggestions? We'd love to hear from you!
Thanks for reading Kiin Bio! Subscribe for free to receive new posts and support my work.
\n
\n\n\n
\n
\n","enclosure":{"link":"https://substack-post-media.s3.amazonaws.com/public/images/598a31a5-8565-4d7a-8293-417288be953c_1200x630.png","type":"image/jpeg"},"categories":[]},{"title":"A Primer on the Comp Bio Career Landscape","pubDate":"2026-06-09 17:01:52","link":"https://newsletter.kiin.bio/p/a-primer-on-the-comp-bio-career-landscape","guid":"https://newsletter.kiin.bio/p/a-primer-on-the-comp-bio-career-landscape","author":"Natasha Kilroy","thumbnail":"","description":"Welcome back to Kiin Bio Weekly.","content":"\n
Welcome back to Kiin Bio Weekly.
\n
Who this piece is for: Computational biology professionals and hiring managers navigating the UK market in 2026.
\n
What this covers: Where the roles are, what they pay, and what actually gets people hired, based on recruiter data and 750+ live listings.
\n
The takeaway: The market rewards specialists who can ship, not generalists who can apply. Infrastructure roles are where demand is highest, entry-level is brutally oversaturated, and your visibility matters more than your credentials.
\n
\n
Freebie alert: We know how hard science is. That\u2019s why we built the Pioneer Programme.
\n
We\u2019re selecting academic and nonprofit teams to get one year of free access to our drug discovery platform, with support from our science team. If you spend more time pulling together findings from different sources than actually acting on them, it\u2019s worth applying.
\n
No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.
The computational biology job market in 2026 has split in two. We wanted to dive deeper to understand why, and try to think about what might happen next.
\n
Traditional pharma has been making considerable cuts for over a year. Bayer alone cut roughly 14,500 roles between 2022 and 2026, with the steepest reductions in 2024 and 2025. BMS and Teva followed with thousands more. AI-native biotech, meanwhile, is hiring faster than the talent pool can seem to keep up. As a quick example, Isomorphic Labs has 21 open ML drug discovery roles in London alone.
\n
We looked into the market dynamics, skill levels, salary benchmarks, and hiring patterns across 2,000+ UK listings on LinkedIn, and spoke to Joe Phillips, Principal Consultant BioAI at Cubiq Recruitment, a specialist recruiter in bio-AI and computational life sciences. Everything we found is in here: where the roles are, what they pay, who\u2019s getting hired, and what we think happens next.
\u201cThere\u2019s an enormous amount of strong academic talent coming through, but the number of genuinely junior opportunities is tiny compared to demand. The candidates who stand out usually have something beyond their degree alone to show.\u201d
\n
\n
\ud83d\udcca Where the roles are
\n
Across the UK, there are currently 2,000+ open positions in computational biology, bioinformatics, AI drug discovery, and adjacent infrastructure roles (source: LinkedIn Jobs, May 2026). These span big pharma, AI-native biotech startups, NHS trusts, CROs, and platform companies selling into life sciences.
\n
\n
MLOps and platform engineering (1,000+ roles): The biggest category by far, and probably the most surprising if you haven\u2019t been watching the infrastructure side of biotech. Why so many? McKinsey\u2019s 2025 State of AI report found that 88% of companies now use AI in at least one business function (up from 78% the year before), but roughly two-thirds are still stuck in pilot mode. Companies built research teams over the last few years, proved that their models work, and now need people who can operationalise them at scale. Joe says this is the biggest shift he\u2019s seen: \u201cA lot of this work was previously absorbed by ML Engineers. Now the cost and complexity around compute, GPU utilisation, and inference has become significant enough that firms are hiring specialists.\u201d
\n
Bioinformatics (355 roles): The broadest category, spanning clinical bioinformatics, genomics, spatial and single-cell analysis. Also the most accessible at entry level: 33% of bioinformatics listings are entry-level, compared to just 12% in ML drug discovery. For early career candidates, this is where the door is most open, across both big pharma and smaller biotech. Startups tend to offer faster progression and broader scope; pharma offers stability and established infrastructure.
\n
Computational biology (164 roles): Core comp bio and adjacent research scientist roles. Spans both pharma and biotech.
\n
ML in drug discovery (137 roles): Protein design, ADMET prediction, molecular modelling. Isomorphic Labs dominates with 21 positions, followed by Relation Therapeutics with 16. Newer players like CuspAI and Boltz have also been hiring heavily over the past 12 months. Small in absolute numbers, but likely one of the fastest-growing categories year on year: the AI in drug discovery market is expanding at around 25% annually (Verified Market Research). Even if the number of roles today looks modest, the trajectory and the funding flowing in suggest this will look very different by 2028. The roles that exist here tend to be senior, well compensated, and competitive.
\n
Clinical AI (~500 roles): The broadest umbrella, covering health informatics through to clinical data science.
\n
AI protein design (9 roles): Tiny, highly specialised, and almost exclusively mid-senior level.
Figure 1. Open roles by sub-field in UK computational biology, May 2026. Data: LinkedIn Jobs.
\n
The other trend Joe flags: forward-deployed engineers and technical client facing hires. \u201cA lot of companies are commercialising scientific ML platforms now rather than running their own therapeutics pipelines, so they need engineers and scientists who can comfortably operate across product, research, and client conversations.\u201d GTM hiring is picking up too as firms move beyond pure research mode.
\n
\n
\ud83d\udccd Geography: the Golden Triangle and beyond
\n
If you\u2019ve been paying attention to the London tech scene, the top of this list won\u2019t surprise you.
\u201cKings Cross is obviously on fire at the moment,\u201d says Joe. \u201cOxford and Cambridge are also hotspots, although of the two, Cambridge is always an easier sell to candidates because of the commutability from London.\u201d
\n
Outside the Golden Triangle, the Northern Arc (Leeds, Liverpool, Manchester, Sheffield) is showing up as a secondary cluster, backed by Northern Gritstone funding for life science and deep tech spinouts. Glasgow also appears consistently in bioinformatics listings, driven by NHS Scotland roles. For candidates willing to look beyond the south-east, the cost of living advantage is real, particularly when London salaries don\u2019t always come with a proportional premium.
\n
\n
\ud83d\udcb0 What it pays
\n
So where do these 2,000+ roles sit on salary? UK data for comp bio is notoriously thin, which makes it hard for candidates to know when an offer is fair. Joe shared benchmarks from his recruitment work:
The ML premium is hard to ignore. A senior ML engineer in biotech can earn nearly double what a senior bioinformatician earns. That comes down to scarcity of strong ML talent and the direct commercial impact these roles carry: GPU optimisation and inference efficiency directly affect a company\u2019s burn rate.
\n
Joe\u2019s caveat: \u201cThis can massively depend on the size of the company and what they are working on. There are always outliers.\u201d Community data backs this up. Biotech equity is less liquid than FAANG RSUs, so even when base salaries match, total compensation often lags tech. The trade-off is mission, ownership, and the fact that senior bio-AI roles are closing the gap faster than any other life sciences category.
\n
\n
\ud83c\udfe0 Remote, hybrid, or on-site?
\n
Many comp bio professionals came up during the pandemic era of fully remote work, so this question comes up constantly. The answer depends on what you actually do day to day.
Bioinformatics is the outlier for remote work (46%), probably because much of the work is pipeline-based and can run independently. Drug discovery and comp bio skew heavily in-person, particularly at early-stage companies where wet lab and dry lab collaboration matters daily.
\n
\u201cFully remote roles are still fairly rare,\u201d says Joe. \u201cIf anything, more companies are trying to get people together in person more, especially at earlier stages where collaboration between science and engineering matters a lot.\u201d
\n
In practice, remote flexibility correlates strongly with sub-field. Bioinformatics and genomics data science offer the most options; ML drug discovery is almost entirely on-site.
\n
\n
\ud83d\udee0\ufe0f Skills that matter now
\n
The roles exist and the salaries are there, the question is really just what actually gets you through the door as the technical bar has moved. Python is the most used programming language in the world (GitHub\u2019s 2024 Octoverse report placed it at number one for the first time, overtaking JavaScript). R remains critical for statistical genomics. Beyond the basics, what actually separates candidates falls into three tiers.
JAX proficiency (driven by the DeepMind/Isomorphic ecosystem)
\n
Production ML deployment (Kubernetes, inference optimisation)
\n
Cross-functional communication: being able to sit across product, research, and client conversations
\n
\n
Emerging (bet on these for 2027):
\n
\n
Diffusion models for molecular generation (e.g. DiffDock, FrameFlow)
\n
Geometric deep learning and equivariant neural networks
\n
LLMs for biomedical data (RAG architectures, agentic AI for research)
\n
Foundation models for genomics (single-cell, spatial transcriptomics)
\n
\n
PyTorch has decisively won over TensorFlow in bio-AI research. JAX has gone from niche to essential for structural biology. Perl has disappeared from modern curricula entirely. The field moves fast enough that what was cutting-edge in 2023 (basic AlphaFold usage) is now baseline knowledge.
\n
What we think: The agentic AI category is the one to watch. Right now, \u201cagentic\u201d is mostly a buzzword on job listings. Within 18 months it will be a real job requirement, because the companies that figure out how to automate their literature review, hypothesis generation, and experimental design pipelines will move significantly faster than those relying on manual researcher effort. If you\u2019re picking a side project to build in public, an agentic research workflow is probably the highest-signal thing you could show a hiring manager right now.
\n
\n
\ud83c\udfaf How to stand out as a company and an employee (from someone who sees 1,000 CVs)
\n
Knowing the right skills is one thing, betting noticed in a pile of 300 applicants is another. The entry-level bottleneck is worth spelling out from both sides. Candidates are applying into a tiny number of junior roles: only 11-12% of ML drug discovery and comp bio positions are entry-level, yet the pipeline of qualified graduates is enormous. Joe puts numbers to it: a senior role might attract 50 applicants, of which maybe one is genuinely relevant. A junior role pulls 300-350, of which 7-10 are a real fit. The ratio of applicants to roles is 6-7x higher at entry level, and even then, most applications miss the mark. Companies, meanwhile, are drowning in applications and still can\u2019t find the right people. The volume of inbound is high; the signal-to-noise ratio is low.
Figure 2. Experience level distribution across comp bio and ML drug discovery roles, UK 2026. Data: LinkedIn Jobs.
\n
Joe\u2019s advice on what separates the top candidates:
\n\n
Show evidence of building, not just researching. \u201cFounders and hiring teams often love seeing evidence that someone can actually build and ship things, not just research them theoretically. Particularly in bio, there\u2019s a real appreciation for candidates who can move between experimentation and execution.\u201d
\n
Be visible where recruiters actually look. This goes beyond LinkedIn. \u201cA huge chunk of our time is tracking publications and collaborations to map who\u2019s working on what, which labs are producing interesting work. There\u2019s also GitHub and Hugging Face, where we\u2019re digging through models, software tooling and tech stacks.\u201d Active repositories, conference contributions, hackathon results, and community involvement all increase visibility.
\n
Go deep in interviews, not broad. \u201cA lot of candidates stay very high-level when explaining their work, but hiring teams usually want to go much deeper than people expect. They want candidates to talk through how decisions were made, why certain approaches worked.\u201d Approach interviews like an external consultant: figure out exactly what you are there to improve.
\n
Bridge the academia-industry gap deliberately. \u201cWhere people sometimes struggle is around product and commercial awareness. Industry teams aren\u2019t just thinking about whether something works academically. They\u2019re thinking about usability, timelines, scalability, deployment and business impact.\u201d The candidates who transition best have already sought out internships, collaborations, or commercial projects while in academia.
\n
Network into the hidden job market. \u201cThe people most in-demand tend to have clear signals that they\u2019re genuinely interested in their area outside their day job. They\u2019re naturally around others in the space a lot of the time, so are closer to that hidden job market that\u2019s often rife with word-of-mouth opportunities.\u201d
\n\n
What we think: The pipeline problem in comp bio is structural, not cyclical. Universities produce graduates faster than the industry can absorb at junior level, while senior roles go unfilled for months. This won\u2019t close by itself. The people who break through are the ones who\u2019ve already demonstrated they can operate above their experience level. A polished GitHub with one well documented, production quality project is worth more than five papers in middling journals. Hiring managers are pattern-matching for \u201ccan this person ship something on day one,\u201d and the evidence needs to be visible before the interview.
\n
\n
\ud83c\udfe2 For companies: how to compete for talent
\n
Now the contrary. If you\u2019re a hiring manager or founder reading this, you already know the challenge. You\u2019re getting hundreds of applications per role and still can\u2019t find the right people. Startups are competing with Isomorphic Labs (\u00a31.6 billion raise in 2025), Google DeepMind, and Boltz for the same small talent pool. Joe\u2019s take on what works:
\n
\u201cA lot of candidates want more ownership, closer access to founders, more influence over direction of a product, and thrive on the ability to actually see their work shape a product directly. The larger players cannot offer this at the same level.\u201d
\n
Smaller companies can lean into that. What the ones that hire well do differently:
\n
\n
Clarity of mission: Strong conviction about what they are building and why it matters. \u201cIf people believe in the founding team and what they\u2019re standing for, it counts for a lot.\u201d
\n
Process efficiency: \u201cSlow feedback, too many stages, or technical tasks that take hours of a candidate\u2019s time can quickly put people off.\u201d Even unsuccessful candidates should leave with a good impression.
\n
Communication throughout: \u201cCompanies sometimes assume that if there\u2019s no update, there\u2019s no reason to contact the candidate. From the candidate\u2019s side, silence feels like being ghosted.\u201d Even an update saying they are still in consideration keeps people engaged.
\n
Clear expectations: \u201cSo often role details change mid-search, sometimes several times, which sends a mixed message to market and damages perception to the target talent pools.\u201d
\n
\n
What we think: The talent competition in bio-AI is asymmetric in a way that favours startups, if they play it right. The big players offer prestige and salary. They cannot offer speed, ownership, or the feeling of shaping something from scratch. The startups that lose candidates to Isomorphic or DeepMind are usually the ones that ran a slow, unclear process, not the ones that lost on compensation alone. Your hiring process is your first product demo. If it\u2019s confusing or inconsistent, strong candidates will read that as a signal about what working there is actually like.
\n
\n
\u26a1 The market in motion
\n
Two things are true at the same time in computational biology right now. Traditional pharma is contracting (patent cliffs, layoffs, restructuring), while AI-native biotech is expanding fast. Isomorphic Labs raised \u00a31.6 billion with no molecules in clinical trials. UK seed investment leapt 19% in 2025. Two UK biotech companies hit unicorn status (Verdiva Bio and Isomorphic Labs). The computational biology market overall is growing at 13% annually toward $22 billion by 2034.
\n
For candidates, the opportunity is real, but \u201clearn Python and apply broadly\u201d doesn\u2019t work anymore. The market rewards specialists who can ship production systems and communicate across disciplines. Infrastructure roles (MLOps, GPU optimisation, deployment) are where the most acute demand is. The salary premium for ML over traditional bioinformatics is widening. Remote work exists, though it\u2019s not the default.
\n
The people who do best in this market aren\u2019t necessarily the most credentialed. They tend to be the ones who are visible, who adapt quickly, and who build things in the open.
\n
\n
\ud83d\udcac Want to be featured in Kiin Bio Weekly?
\n
Each issue we speak directly with researchers, scientists, and builders working at the frontier of AI in life sciences. If you're working on something in this space and think it would resonate with our community, I'd love to hear from you. Fill out this form or reach out to me directly.
\n
\n
Found this useful? Forward it to a colleague in computational biology, it's the best way to help the newsletter grow.
Subscribe now to stay at the forefront of AI in Life Science. Every week: primers, deep dives, and direct conversations with the people building the field.
Welcome back to your weekly dose of AI news for Life Science!
\n
Three papers this week that all ask some version of the same question: are we actually measuring the right thing? SwitchCraft pushes protein design past the single-structure assumption that underpins most current methods. TadA-Bench asks whether protein language models can predict the future of evolution or only interpolate its past. Chem-PerturBridge harmonises the messy pile of perturbation transcriptomics data and reveals just how little of it agrees at the gene level. The connecting thread: the foundation models exist, but the design paradigms and benchmarks haven\u2019t caught up.
\n
\n
We just opened up our Kiin Pioneer Programme: free access to our platform for academic and nonprofit research teams for a year.
\n
The short version: we\u2019ve built a place where scientists can collaborate on their drug discovery work without everything living in disconnected tools and someone\u2019s local files. Literature reviews, target discovery, bioinformatics, all in one place. When one person finds something interesting and someone else has relevant data, the platform catches that and suggests what to look at next. Everything\u2019s tracked, so six months later you actually know what was done and why.
\n
We\u2019re looking for teams who are trying to move faster on questions like: which targets should we prioritise? How do we make sense of conflicting evidence? What\u2019s actually worth testing next?
\n
No cost, no data transfer, all IP stays with your institution. Applications close in August, first cohort starts in September.
\ud83d\udd2c Protein design tools optimise for one structure. Real proteins switch between conformations to do their jobs: allosteric enzymes toggle between active and inactive states, biosensors change shape upon binding. Designing proteins that intentionally switch between defined states has been a manual, low-throughput exercise.
\n
Jing, Bafna, and colleagues from MIT\u2019s Berger lab built SwitchCraft, a framework that designs proteins with specified multi-state behaviour by backpropagating through differentiable structure prediction models.
\n
\ud83e\uddec The core idea: treat Boltz-1 (an open-source structure prediction model) as a differentiable loss function. Define the desired structural states, then optimise the sequence so that it folds into all of them under appropriate conditions. The framework is programmatic. You compose design objectives from modular building blocks rather than training a new model per task.
\n
\u26a1 They demonstrate allosteric regulation of protein motifs, discrimination between bound ligand identities, and fluorescent biosensor design. The biosensor results are particularly telling: designed sequences show distinct fluorescence states depending on which ligand is bound. That\u2019s functional multi-state behaviour that hasn\u2019t been accessible through standard single-state design pipelines.
This sits at the leading edge of computational protein design, downstream of structure prediction but upstream of experimental characterisation. The reason it exists now is simple: differentiable structure prediction. Until models like Boltz-1, ESMFold, and AlphaFold became fast and accurate enough to use as gradient-providing modules, you couldn\u2019t backpropagate through structure. RFdiffusion and ProteinMPNN design single structures beautifully, but they don\u2019t handle the multi-state problem. SwitchCraft does something conceptually different: it treats structure prediction as a subroutine rather than the end goal. The limitation is that experimental validation here is still computational, relying on Boltz-1\u2019s predictions of the designed states. Whether these sequences actually fold into multiple states in a wet lab remains open, though the biosensor designs are testable. If you work on biosensors, allosteric switches, or molecular logic gates, this is worth trying now.
\n
\ud83d\udca1 Why This Is Cool
\n
This is what it looks like when structure prediction becomes infrastructure rather than the main event. The field spent five years building accurate folding models. Now those models are components in design loops. Multi-state protein design has been a goal since the Kuhlman lab\u2019s early work on conformational switches, but the computational tools never matched the ambition. SwitchCraft doesn\u2019t solve the full problem (experimental validation is still the bottleneck), but it makes the design step tractable in a way it simply wasn\u2019t before.
\ud83d\udd2c Protein fitness prediction benchmarks typically test whether models can fill in gaps within a mutational landscape. That\u2019s interpolation. What protein engineers actually need is extrapolation: given rounds 1 through 10 of directed evolution, can you predict what we\u2019ll find useful in round 11? No existing benchmark properly tests this.
\n
Gao and colleagues from Shanghai Jiao Tong University built TadA-Bench from 31 rounds of real TadA (tRNA adenosine deaminase) directed evolution, totalling roughly one million variants.
\n
\ud83e\uddec The benchmark enforces chronological evaluation: models train on earlier rounds and must predict which variants from later rounds will be experimentally validated. It provides aligned DNA, RNA, and protein sequences, and uses a Seq2Graph method to create comparable activity measurements across rounds that originally used different assay conditions.
\n
\u26a1 The headline finding is sobering. Current protein language models (including ESM-2 and various fine-tuned variants) struggle with temporal prediction even when they perform well on standard interpolation benchmarks. The gap between interpolation and extrapolation performance is substantial. Good performance on DMS datasets doesn\u2019t mean your model can guide the next round of experiments. Many practitioners suspected this; now there\u2019s a number attached to it.
This is a benchmark, not a tool, and its value depends on adoption. It fills a gap the field has known about for years. ProteinGym and similar resources test interpolation well, but nobody had assembled a large-scale temporal benchmark from real directed evolution campaigns. The 31 rounds of TadA evolution provide unusually rich temporal structure (most published datasets have 3-5 rounds at most). The \u201cagentic protein engineering\u201d framing is timely: as more groups wire LLMs into experimental design loops, you need evaluation frameworks that test whether the model\u2019s suggestions actually lead somewhere productive. The dataset is on Hugging Face and the code is open on GitHub, which lowers the barrier to adoption. The main caveat is generalisability: TadA is one enzyme. Performance on this benchmark won\u2019t guarantee performance on your protein of interest. Still, it\u2019s the best temporal test we have.
\n
\ud83d\udca1 Why This Is Cool
\n
The protein ML field has a measurement problem. Papers report NDCG on held-out DMS positions and claim their model \u201cguides protein engineering.\u201d This benchmark calls that bluff. Can your model predict the future, or only reconstruct the past? The distinction matters enormously for anyone running a real directed evolution campaign. If current models fail at temporal prediction (and they largely do), that\u2019s uncomfortable but useful information. It tells you where the actual research gap is, which is more valuable than another leaderboard.
\ud83d\udd2c The field has generated enormous amounts of small molecule perturbation transcriptomics data (L1000, sci-Plex, Tahoe, and many smaller datasets). The problem: nobody knows how well they agree with each other, and combining them for model training requires harmonisation that hasn\u2019t been done systematically.
\n
Sza\u0142ata and colleagues from the Theis lab at Helmholtz Munich built Chem-PerturBridge, standardising 37,000+ compounds across 1.25 million samples from eight assay types with consistent metadata.
\n
\ud83e\uddec The compendium spans nine datasets including sci-Plex3, Tahoe, L1000, OP3, DILImap, and others. The harmonisation pipeline normalises cell line annotations, compound identifiers, and gene nomenclature, making cross-dataset comparison possible for the first time at this scale.
\n
\u26a1 The uncomfortable finding: gene-level agreement between datasets is poor. When the same compound is tested in the same cell line across different platforms, correlation at the individual gene level is low. Directional consistency (is the gene up or down?) is better, and turns out to be sufficient for improving compound representation learning. Models trained on the harmonised directional data outperform those trained on any single dataset alone
This sits upstream of almost everything in computational chemical biology: virtual screening, target identification, mechanism-of-action prediction, and toxicity forecasting all depend on perturbation data quality. The honest finding here is that the perturbation transcriptomics field has a reproducibility problem at the gene level, even between high-quality experiments. That\u2019s not new as a suspicion, but quantifying it across eight assay types is valuable. The positive takeaway is that directional signals persist, and they\u2019re enough to learn useful compound representations. For practitioners, this means you should probably stop training on raw gene-level expression values from a single perturbation study and start using directional features from multiple sources. The resource is open (MIT licence on code, upstream licences on data), which is appropriate for something positioned as community infrastructure.
\n
\ud83d\udca1 Why This Is Cool
\n
The field has been treating perturbation transcriptomics datasets as interchangeable training data without checking whether they actually say the same thing. They don\u2019t, at least not at the resolution most people assume. This paper does the unglamorous work of quantifying that disagreement and showing what signal does survive. The practical consequence: if you\u2019re building models on LINCS L1000 data alone, you\u2019re probably leaving performance on the table. Directional agreement across platforms is a more robust training signal than absolute expression values from one platform. This is infrastructure work rather than a methods advance, but it\u2019s the kind that makes everything downstream more trustworthy.
The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.
A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. \u00a360 for biotech companies.
The first edition of a one-day symposium bringing together London\u2019s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub L\u00e1la. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.
ELIXIR\u2019s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that\u2019s your window.
\n
\n
Thanks for reading!
\n
\ud83d\udcac Get involved
\n
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Connect With Us
\n
Have questions or suggestions? We'd love to hear from you!
Welcome back to your weekly dose of AI news for Life Science!
\n
I keep coming back to this question of whether the protein ML field has a modelling problem or a data problem. This week's papers make the case from both sides. Biohub's ESM release argues that scale and architecture can solve binder design from scratch. OpenADMET argues the opposite: that no model will work until the training data stops being inconsistent rubbish. SynFit sits somewhere in between, showing what you can get when you have good multi-property data and a framework that knows how to use it.
\n
\n
We just launched our Kiin Pioneer Programme, giving academic and nonprofit research teams one year of free access to our drug discovery platform!
\n
KiinOS is a platform where scientists can run literature reviews, target discovery, and bioinformatics in one place. It keeps a record of what\u2019s been done, by who, and what came out of it. So if one person finds a promising target and someone else has relevant data, the platform connects those results and suggests what to pursue next.
\n
We\u2019re looking for teams asking: which targets should we prioritise? How do we interpret conflicting evidence? Which hypotheses are worth testing next?
\n
There\u2019s no cost, no data transfer, and all IP stays with your institution. Applications close in August, with the first cohort starting September.
\ud83d\udd2c We can predict how proteins fold, but designing new ones that actually bind specific targets, particularly antibodies, still requires months of expensive lab screening just to find starting candidates.
\n
Biohub has released ESM: a protein language model (ESMC, trained on 2.8 billion sequences), a structure prediction and design model (ESMFold2), and a map of 6.8 billion sequences (ESM Atlas). All MIT-licensed.
\n
\ud83e\uddec ESMFold2 learns protein representations from evolutionary data, then searches that learned space for proteins predicted to bind a given target. It scores candidates using its own confidence estimates, so the entire design loop is computational. Structure prediction runs from a single sequence without needing alignment databases.
\n
\u26a1 Designed binders for five cancer/immunology targets were validated in the lab: minibinder success rates of 70%, scFv antibody success rates of 21%. A PD-L1 binder hit 4.3 nM affinity and blocked immune checkpoint suppression in cells. Cryo-EM confirmed an EGFR binder matched the prediction at 1.2 \u00c5 RMSD.
This positions ESM as a competitor to both AlphaFold 3 (for structure prediction) and RFdiffusion/ProteinMPNN (for binder design). The antibody-antigen prediction results are interesting because this is where AlphaFold has historically struggled most. The binder design piece is where Biohub is making their boldest claim: that the earliest stage of therapeutic protein discovery can happen computationally in days rather than months. Worth noting that the 21% scFv success rate, while a step up from near-zero for most computational methods, still means roughly 80% of designs fail in the lab. These are first-round hits. David Baker\u2019s group (RFdiffusion) and Generate Biomedicines (Chroma) are the obvious comparisons for generative protein design. ESMFold2 is differentiated by working from a language model backbone rather than a diffusion architecture, which changes how it scales with compute at inference time.
\n
\ud83d\udca1 Why This Is Cool
\n
The sparse autoencoder analysis is what I find most thought-provoking. ESMC independently recovered biological concepts like the nucleophilic elbow motif across 75 of 99 relevant enzymes, despite never being told what one is. That\u2019s a model learning the grammar of protein biology from sequence alone. Whether the binder design numbers hold up across a broader and more difficult target set remains to be seen. The cryo-EM validation is reassuring, but five targets is still five targets.
\ud83d\udd2c Protein engineering almost always requires optimising multiple properties at once, but current ML fitness predictors handle each property independently. Train separate models for yield and selectivity and you\u2019ll get variants that excel at one while tanking the other.
\n
Georgia Tech and UC Santa Barbara developed SynFit, a multi-objective framework that fine-tunes protein language models on experimental fitness data across multiple assays simultaneously.
\n
\ud83e\uddec SynFit combines a shared ESM2 encoder with property-specific prediction heads, using contrastive learning to capture cross-property relationships from deep mutational scanning data. Predictions are integrated via Pareto sorting to find variants that improve everything at once.
\n
\u26a1 On Pareto front analysis across 20 proteins, SynFit hits the optimal front 70% of the time versus 60% for ProteinNPT and 55% for ConFit. The wet-lab result is more convincing: 83 out of 100 designed hextuple mutants for a biocatalytic borylation enzyme showed simultaneously improved yield and enantioselectivity, with multiple variants beating everything in the training data.
This sits squarely in the protein engineering workflow after you have initial DMS data and want to explore combinatorial sequence space. ConFit, from the same group, is the direct predecessor, and SynFit extends it with the multi-objective training component. ProteinNPT (Notin et al., ICML 2023) tackles a related problem through non-parametric transformers but doesn\u2019t explicitly model cross-property correlations. The wet-lab validation on biocatalytic borylation is well chosen because it\u2019s a genuinely new-to-nature reaction where directed evolution data is sparse, making the Pareto optimisation problem harder. The limitation is that you still need initial multi-property DMS data for your protein of interest, which isn\u2019t always available. The KRAS case study (identifying shared functional residues across six binding partners) is a nice mechanistic demonstration but relies on an unusually well-characterised system.
\n
\ud83d\udca1 Why This Is Cool
\n
The 83/100 result is more impressive than any of the benchmark numbers. Getting computationally designed hextuple mutants to simultaneously beat the training set on both yield and enantioselectivity, in a single round without iterative screening, is a practical result that enzyme engineers will care about. It suggests ML-guided combinatorial design can start to compress the \u201cdesign-build-test\u201d cycle for multi-objective problems. The architecture is straightforward enough that adoption shouldn\u2019t be difficult once the code is released.
\ud83d\udd2c Around 30% of clinical drug failures trace back to ADMET problems. The ~100 proteins responsible (CYPs, hERG, transporters, nuclear receptors) are well known, but existing ML models train on data cobbled from dozens of labs using different protocols. A recent analysis found almost no correlation between IC50 values for the same compound measured by different groups.
\n
Fraser (UCSF), Edgar (Octant), Chodera (MSKCC), and Walters (OMSF) have launched OpenADMET, an ARPA-H and Gates Foundation-funded consortium generating systematic, internally consistent ADMET datasets and releasing everything publicly.
\n
\ud83e\uddec The consortium runs assays across the full \u201cAvoid-ome\u201d panel at industrial scale: 30,000 compounds per run in 1536-well plates, under $0.40 per compound. Active learning selects informative compounds for expansion, and structural biology (100+ PXR crystal structures so far) resolves binding modes.
\n
\u26a1 This is a programme-level perspective paper rather than a single dataset release. The first community challenge has run. They\u2019re screening tens of thousands of compounds weekly. I appreciate the honesty that this is long-term infrastructure. They\u2019re not claiming to have solved ADMET prediction; they\u2019re arguing nobody will until the data problem is addressed.
This is an upstream data-generation effort, not a prediction tool. It sits before everything else in the ADMET pipeline: the datasets it produces will train the next generation of models. TDC (Therapeutics Data Commons) and ChEMBL aggregate existing literature data. OpenADMET\u2019s bet is that literature data is too inconsistent to train reliable models, and that generating internally consistent measurements from scratch, with structural validation, justifies the cost. Tools like ADMET-AI (Swanson et al., 2024) would be downstream consumers of these datasets. The federated learning alternative (training behind pharma company firewalls) is explicitly discussed and dismissed: it can\u2019t generalise beyond local chemical space and doesn\u2019t produce the structural understanding needed for true mechanistic models.
\n
\ud83d\udca1 Why This Is Cool
\n
The framing is what matters here. By defining ADMET as a finite structural biology problem (map the interactions with roughly 100 proteins and you\u2019ve covered most failure modes), they turn an open-ended prediction challenge into a bounded experimental campaign. The question is whether $0.40-per-compound assays and active learning can produce models that generalise to novel chemical matter outside the training distribution. The open-science commitment, with Gates Foundation backing, makes this more credible than most \u201cwe\u2019ll share data eventually\u201d promises from pharma-adjacent initiatives.
The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.
A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. \u00a360 for biotech companies.
The first edition of a one-day symposium bringing together London\u2019s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub L\u00e1la. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.
ELIXIR\u2019s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that\u2019s your window.
\n
\n
Thanks for reading!
\n
\ud83d\udcac Get involved
\n
We\u2019re always looking to grow our community. If you\u2019d like to get involved, contribute ideas or share something you\u2019re building, fill out this form or reach out to me directly.
\n
Connect With Us
\n
Have questions or suggestions? We'd love to hear from you!
Thanks for reading Kiin Bio! Subscribe for free to receive new posts and support my work.
\n
\n\n\n
\n
\n","enclosure":{"link":"https://substack-post-media.s3.amazonaws.com/public/images/83eceabf-2df2-48c8-833c-a76886f2d6f8_1200x630.png","type":"image/jpeg"},"categories":[]},{"title":"A Primer on Clinical AI \ud83c\udfe5","pubDate":"2026-05-26 17:01:55","link":"https://newsletter.kiin.bio/p/a-primer-on-clinical-ai","guid":"https://newsletter.kiin.bio/p/a-primer-on-clinical-ai","author":"Natasha Kilroy","thumbnail":"","description":"Close to 1,500 FDA-approved AI use cases, meta-analyses proving cost-effectiveness, and hospitals still aren't adopting.","content":"\n
Welcome back to Kiin Bio Weekly.
\n
This week we\u2019re looking at clinical AI, specifically the type that nobody\u2019s talking about. I got connected to Ignacio through his work at Savana, where they\u2019ve been extracting real-world evidence from clinical records for years. What struck me the most was how much of the conversation around AI in medicine misses the point: everyone\u2019s focused on ChatGPT and scribes, while the predictive models that actually enable personalised medicine are sitting there with regulatory approval and population-level evidence, largely unused.
\n
We got on a call, and the result is this primer.
\n
\n
Freebie alert: We know how hard science is. That\u2019s why we built the Pioneer Programme.
\n
We\u2019re selecting academic and nonprofit research teams to get one year of free access to our drug discovery platform plus hands-on support from our science team. If your research bottleneck isn\u2019t data but connecting the findings you already have, this is for you.
\n
We\u2019re looking for teams asking: which targets should we prioritise? How do we interpret conflicting evidence across datasets? Which hypotheses are worth testing next? Where are the strongest translational opportunities?
\n
No cost. No data transfer. All IP stays with your institution. Applications close August, cohort starts September.
Healthcare systems across Europe are breaking. Waiting lists stretch to 14 months. The UK Health Secretary declared the NHS \u201cbroken.\u201d Spain is close behind. Populations are ageing, medicines are expensive, and the workforce cannot scale. Into this crisis arrives artificial intelligence, not as a future promise, but as a present reality. The FDA has approved close to 1,500 AI use cases across clinical specialties. Meta-analyses now demonstrate cost-effectiveness in diabetes, colon cancer, and mammography screening. The question is no longer whether AI works in medicine. It is why adoption remains so slow.
\n
Yet most clinicians, when asked about AI, will talk about ChatGPT and medical scribes. They are looking at the louder revolution while the quieter, more consequential one unfolds beneath it.
\n
We spoke to Ignacio H. Medrano, neurologist-turned-CEO of Savana, about the real state of clinical AI: what is proven, what is hype, and what clinicians are getting wrong about the pace of change.
\u201cIt\u2019s unethical not to use AI these days in certain areas, because it\u2019s proven, and it\u2019s proven at scale across populations.\u201d
\n
\n
\ud83d\udd00 Two types of AI, one common misunderstanding
\n
The most pervasive misconception among clinicians today is that generative AI (LLMs, chatbots, multi-modal models) is the only AI that matters. It is visible, fast-moving, and immediately useful: scribes that eliminate documentation burden, literature tools like Open Evidence replacing PubMed searches, agents managing waiting lists. These applications are exploding because they save time and, critically, do not require clinical validation. They handle documentation, not decisions.
\n
But the deeper disruption is discriminative AI: predictive, classification-based models that have existed for over a decade. This is the AI that enables precision medicine. Granular predictions for individual patients about immunotherapy response, relapse probability in multiple sclerosis, optimal drug sequencing in haematologic cancers. It takes statistics to a level where you can determine the actual probability of a specific outcome for a specific patient.
\n
Discriminative AI started earlier but arrives later in practice. Every algorithm requires validation, external replication, meta-analysis, and integration into clinical guidelines. That pipeline is slow. But it is the pipeline that delivers personalised medicine, and it is now producing results.
\n
\u201cA big misconception is forgetting that discriminative, predictive AI is the real silent disruption,\u201d says Medrano. \u201cThe other is underestimating the speed at which agentic AI is arriving. People think this takes 20 years. It\u2019s happening now, like thunder.\u201d
\n
\n
\u2705 Separating signal from noise: what makes clinical AI \u201creal\u201d
\n
With close to 1,500 FDA-approved AI applications and an exponential curve of machine learning publications on PubMed, distinguishing proven tools from hype requires a framework.
Figure 1. Distribution of FDA-cleared AI/ML medical devices by specialty (cumulative through end-2025). Radiology accounts for three-quarters of all approvals, with cardiovascular medicine a distant second.
\n
Medrano uses three levels:
\n\n
Regulatory approval: FDA or EMA clearance confirms correct dataset construction, generalisation to new cohorts, and absence of bias. This is the minimum threshold.
Figure 2. Kaplan-Meier survival curves from a pragmatic RCT of 15,965 patients. AI-enabled ECG alerts reduced all-cause mortality (HR 0.83, p=0.040), with the strongest effect in AI-identified high-risk patients (HR 0.55, p=0.006). Lin et al., Nature Medicine, 2024.
\n
Real-world deployment: Hospitals actually running these systems in clinical workflows. Hospital Clinic Barcelona has used AI to predict sepsis in intensive care for over three years. Finland deploys predictive models on GP workstations for population segmentation. In China, Ping An Good Doctor attends 100 patients daily without human involvement. In Utah, AI autonomously renews prescriptions.
\n
The gap between levels two and three, between proven effectiveness and actual deployment, is where the real problem lives.
\n
\n
\ud83d\udea7 Three barriers, one that matters most
\n
Technical: Clinical information remains fragmented across systems. Standards like HL7 FHIR are improving interoperability, but the problem is not fully solved. Training and validating models still requires stitching together disparate data sources.
\n
Regulatory: Largely resolved. European regulation now recognises that anonymised data used for research does not require individual informed consent. The European Health Data Space will mandate hospitals to share data within one to two years. Countries like Switzerland, France, Germany, and the UK already permit compliant secondary use of clinical data.
\n
Cultural: The real bottleneck. Not rejection; most managers and clinicians accept AI is inevitable. The problem is twofold. First, pure ignorance: hospital managers do not realise they could build pharmacogenomic models today that predict which patients will respond to expensive biologics, saving millions by paying only for drugs that will work. Second, politics: an algorithm that reduces the need for human labour is nearly impossible to sell when unions and the public demand 10 new nurses before any technology investment. Politicians choose nurses even when they understand AI\u2019s value.
\n
\u201cIt\u2019s not belief anymore, it\u2019s knowledge,\u201d says Medrano. \u201cManagers just don\u2019t understand the power of what\u2019s already possible with the data they have.\u201d
\n
\n
\ud83d\udcca Real-world evidence: from luxury to necessity
\n
Real-world evidence has always mattered. It is the difference between reading about a country and landing there. Clinical trials tell you what should happen under controlled conditions. Real-world evidence tells you what actually happens.
\n
The barrier was always collection. Patient by patient, variable by variable, manually assembling registries over years. Exhausting, expensive, and therefore underutilised. Now, computational systems can extract this information automatically, reliably, and at scale from electronic health records. Once extraction became feasible, demand exploded. Regulators began requesting it. Pharma began requiring it.
\n
This is where initiatives like the UK\u2019s Foresight programme become significant: 57 million medical records feeding predictive models for 100 diseases at 20-year horizons. The Scandinavian countries (Norway and Denmark) are sharing data internationally and validating models across borders. These are not pilots. They are national-scale infrastructure decisions.
\n
\n
\ud83e\udd16 The convergence: agentic AI meets predictive models
\n
Here is where the field is heading, and where most clinicians have not yet looked. The users of sophisticated discriminative AI models (multi-modal predictive algorithms trained on clinical text, genomics, proteomics, radiomics) will not be human doctors. They will be certified agentic AI systems.
\n
Generative AI agents will orchestrate clinical workflows. Discriminative AI will provide the predictions those agents act on. The agent decides what to ask; the predictive model provides the answer. The human clinician supervises, validates, and handles what requires physical presence.
\u201cIf doctors keep doing the same thing, they\u2019ll become pointless,\u201d says Medrano. \u201cPeople will turn to their phones, where an agentic AI healthcare service for 10 euros will give them advice, pull their data, run algorithms. And then that agent will hire humans to perform the physical tasks it cannot.\u201d
\n
\n
\u26a1 Where this leaves us
\n
The infrastructure is being built. Data-sharing mandates are arriving. Validation evidence is accumulating. The two types of AI, generative and discriminative, are converging toward agentic systems that will reshape how healthcare is delivered.
\n
The bottleneck is not technology or regulation. It is the speed at which institutions recognise what is already possible, and act before the system breaks entirely, or before patients simply route around it.
\n
Big thanks Ignacio for meeting with us and sharing his insights for this primer!
\n
\n
\ud83d\udcac Want to be featured in Kiin Bio Weekly?
\n
Each issue we speak directly with researchers, scientists, and builders working at the frontier of AI in life sciences. If you're working on something in this space and think it would resonate with our community, I'd love to hear from you. Fill out this form or reach out to me directly.
\n
\n
Found this useful? Forward it to a colleague in drug discovery or computational chemistry, it's the best way to help the newsletter grow.
Subscribe now to stay at the forefront of AI in Life Science. Every week: primers, deep dives, and direct conversations with the people building the field.
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!cC_Y!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F972884f4-5c43-45ee-9411-533d417ef8a7_1920x1080.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!cC_Y!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F972884f4-5c43-45ee-9411-533d417ef8a7_1920x1080.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!446I!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F02b133af-4817-4016-a3c0-da31d46eb4dc_1756x908.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!DCxe!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F81744a90-d4c2-4ed1-9239-2d2ff2442ade_881x905.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!pg13!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F702e82ed-b4e6-4b55-8284-0c95d6b8ded1_697x826.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!FO84!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc5f4723c-5135-461d-ba51-30e9483a121e_1750x953.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!1_vF!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ba1dc64-7f54-47ff-ae71-d24a5e0cca4d_800x800.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!vF-B!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3b3d61-21da-4526-94a1-a8becf48d3fc_1536x1024.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!rqW5!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F18cf985d-518f-4a20-a479-24337b762b8f_840x346.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!IKUp!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F78ff91c7-0328-42b1-813b-2460e82dd650_940x204.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!l8f-!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7fe6f2cf-d26e-4be3-a938-ad7ca24f65ed_1282x1286.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!xGDb!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0de841e9-2cf7-4382-9f87-41309cf28c21_800x800.jpeg\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!I-Vf!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F80cbf68f-7ae9-4294-978a-8c63cad2f699_1200x1200.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!a5qw!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F30f2016f-cf01-491b-9a20-92eadc31d307_1200x700.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!rK2c!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffa33a0da-5d39-4c64-8890-4485b5b80ce5_1200x580.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!m-i8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff1198037-51e4-4b74-8721-6ce396a431f3_1200x660.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!J6GL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd61330-8598-4e1a-9b00-4c0b431e05ce_1200x1200.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!CEp7!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3a771c7-7a5e-4661-83b4-093a8ae37bd2_2048x1210.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!s1zF!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa09f9d6b-00cb-4668-93c5-537f6485e3c8_1822x748.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!tzSz!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fb06e8a0f-ffe7-4384-9b67-185c2d7ac036_1750x542.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!VR7i!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1632b4d8-0784-4628-8b21-c99fc02986d3_2154x1212.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!IPcm!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Feff9b7ea-2476-46dd-a705-09a116c7039f_1604x1410.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!eiNh!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F01915b91-d0f3-4e97-bf40-ec4ad2632558_1200x1200.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!KKyH!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa31dbf84-3578-4008-ae0d-b35b5855821a_1954x880.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!Ja3o!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e9e25d9-f383-422b-b1c5-eb5227b34f58_1766x1190.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!F13Y!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1aeb7b08-c862-46bf-97f9-186a45144400_1586x1260.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!xUdJ!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fef45c606-d447-4f88-9a7e-6098739ba7eb_3840x2160.png\" "\\"\\"")
![\"Fig.](\"https://substackcdn.com/image/fetch/%24s_!cji1!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcbba7592-20aa-4dbf-936b-08e021fb4eda_2168x1517.png\" "\\"Fig.")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!uNrz!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5cf6be95-cea0-4844-adf4-33fef031038b_2012x922.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!fLzS!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6cd5a428-68ae-4574-9cb2-44bc020160d1_1802x594.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!Y4_W!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fdc034ace-cd66-46f5-a43a-5d23c613bd0d_1582x1172.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!Unge!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fae3bad2f-a8f4-4779-a765-f20b80bb9de6_1724x902.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!BScV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F4c7ff958-6bab-4167-8ce4-0fdd2cbcd683_2040x1220.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!0kH3!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7447d310-eeb7-4db7-a9b0-f0962bef897f_1890x874.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!vY7P!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa13e3ffb-0f3c-4a9e-a1ce-144cba498573_1872x1256.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!Yi6K!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6de3163b-4e76-4ee0-b3d2-8a643ccc43a6_1526x1356.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!c6c9!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F657ec9f9-3f8a-4739-ac8e-36ccd9aa221b_1872x1282.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!3gVq!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fb8f85d77-2f87-4223-b9e1-7351f59b68ec_2052x1166.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!q1BS!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc0aae58b-af30-445e-9ad2-5edd6810a23f_2020x940.png\")
![\"\"](\"https://substackcdn.com/image/fetch/%24s_!1P2Q!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F179a77dd-6bb5-41a6-98a9-7deb2d23e9b7_1365x2048.jpeg\")