Tudor-Stefan Cotet
@tudorcotet.bsky.social
MSc Biotechnology student at ETH Zürich | ESOP Scholar 2023 | Computational immunology and protein design enthusiast
Reposted by Tudor-Stefan Cotet
Today we're dropping the "beta" tag from Adaptyv, launching our new website and announcing our $8M seed round.
When we started Adaptyv a few years ago, our core belief was: AI models for biology are only as good as the lab data they're trained on and the hypotheses they can test in the real world.
When we started Adaptyv a few years ago, our core belief was: AI models for biology are only as good as the lab data they're trained on and the hypotheses they can test in the real world.
September 18, 2025 at 5:40 PM
Today we're dropping the "beta" tag from Adaptyv, launching our new website and announcing our $8M seed round.
When we started Adaptyv a few years ago, our core belief was: AI models for biology are only as good as the lab data they're trained on and the hypotheses they can test in the real world.
When we started Adaptyv a few years ago, our core belief was: AI models for biology are only as good as the lab data they're trained on and the hypotheses they can test in the real world.
Reposted by Tudor-Stefan Cotet
We compared the calibration of various machine learning uncertainty estimation methods for protein engineering.
No method excels across all scenarios, and uncertainty-based strategies for optimization often did not outperform methods without uncertainty.
No method excels across all scenarios, and uncertainty-based strategies for optimization often did not outperform methods without uncertainty.
January 9, 2025 at 1:31 PM
We compared the calibration of various machine learning uncertainty estimation methods for protein engineering.
No method excels across all scenarios, and uncertainty-based strategies for optimization often did not outperform methods without uncertainty.
No method excels across all scenarios, and uncertainty-based strategies for optimization often did not outperform methods without uncertainty.
Reposted by Tudor-Stefan Cotet
Cryptic pockets are hidden binding sites in proteins that are not visible in their standard ("apo") structure and only stable in the presence of the right ligand.
Here, Bemelmans et al. review how scientists use computer algorithms to detect them.
www.sciencedirect.com/science/arti...
🧪 #CompChem
Here, Bemelmans et al. review how scientists use computer algorithms to detect them.
www.sciencedirect.com/science/arti...
🧪 #CompChem
Computational advances in discovering cryptic pockets for drug discovery
A number of promising therapeutic target proteins have been considered “undruggable” due to the lack of well-defined ligandable pockets. Substantial r…
www.sciencedirect.com
January 8, 2025 at 1:00 PM
Cryptic pockets are hidden binding sites in proteins that are not visible in their standard ("apo") structure and only stable in the presence of the right ligand.
Here, Bemelmans et al. review how scientists use computer algorithms to detect them.
www.sciencedirect.com/science/arti...
🧪 #CompChem
Here, Bemelmans et al. review how scientists use computer algorithms to detect them.
www.sciencedirect.com/science/arti...
🧪 #CompChem
Reposted by Tudor-Stefan Cotet
January 3, 2025 at 5:39 PM
Reposted by Tudor-Stefan Cotet
Blown away by this new paper from the King lab: "Hierarchical design of pseudosymmetric protein nanocages"
www.nature.com/articles/s41...
www.nature.com/articles/s41...
December 21, 2024 at 2:39 AM
Blown away by this new paper from the King lab: "Hierarchical design of pseudosymmetric protein nanocages"
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Reposted by Tudor-Stefan Cotet
A protein's position in the genome in relation to other genes is informative about function, so a genomic language model can be prompted to generate
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
December 18, 2024 at 10:14 PM
A protein's position in the genome in relation to other genes is informative about function, so a genomic language model can be prompted to generate
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
- toxin/antitoxin pairs
- anti-CRISPR proteins
+ a database of 120B synthetic base pairs
www.biorxiv.org/content/10.1...
@brianhie.bsky.social
Reposted by Tudor-Stefan Cotet
Paper #1: arxiv.org/abs/2412.12979
Aligning autoregressive pLM's to generate EGFR binders via Direct Policy Optimization (DPO) from the incredible @noeliaferruz.bsky.social who gave a great talk as part of the MLSB workshop
Aligning autoregressive pLM's to generate EGFR binders via Direct Policy Optimization (DPO) from the incredible @noeliaferruz.bsky.social who gave a great talk as part of the MLSB workshop
Guiding Generative Protein Language Models with Reinforcement Learning
Autoregressive protein language models (pLMs) have emerged as powerful tools to efficiently design functional proteins with extraordinary diversity, as evidenced by the successful generation of divers...
arxiv.org
December 20, 2024 at 4:42 PM
Paper #1: arxiv.org/abs/2412.12979
Aligning autoregressive pLM's to generate EGFR binders via Direct Policy Optimization (DPO) from the incredible @noeliaferruz.bsky.social who gave a great talk as part of the MLSB workshop
Aligning autoregressive pLM's to generate EGFR binders via Direct Policy Optimization (DPO) from the incredible @noeliaferruz.bsky.social who gave a great talk as part of the MLSB workshop
Reposted by Tudor-Stefan Cotet
I’m so excited to share that some of my grad work in the @aaronwhiteley lab is now available on bioRxiv! We used an AlphaFold-multimer screen to uncover how a bacterial NLR-related proteins senses multiple phage-encoded proteins to confer phage protection. Link: www.biorxiv.org/content/10.1...
A bacterial NLR-related protein recognizes multiple unrelated phage triggers to sense infection
Immune systems must rapidly sense viral infections to initiate antiviral signaling and protect the host. Bacteria encode >100 distinct viral (phage) defense systems and each has evolved to sense cruci...
www.biorxiv.org
December 18, 2024 at 8:08 PM
I’m so excited to share that some of my grad work in the @aaronwhiteley lab is now available on bioRxiv! We used an AlphaFold-multimer screen to uncover how a bacterial NLR-related proteins senses multiple phage-encoded proteins to confer phage protection. Link: www.biorxiv.org/content/10.1...
Reposted by Tudor-Stefan Cotet
(1/5) Traditionally, antibiotic discovery has centered on bacteria and fungi—until now.
For the first time, we present a systematic exploration of Archaea—a major yet underexplored branch of the tree of life—as a source of novel antimicrobial compounds. www.biorxiv.org/content/10.1...
For the first time, we present a systematic exploration of Archaea—a major yet underexplored branch of the tree of life—as a source of novel antimicrobial compounds. www.biorxiv.org/content/10.1...
December 12, 2024 at 10:56 AM
(1/5) Traditionally, antibiotic discovery has centered on bacteria and fungi—until now.
For the first time, we present a systematic exploration of Archaea—a major yet underexplored branch of the tree of life—as a source of novel antimicrobial compounds. www.biorxiv.org/content/10.1...
For the first time, we present a systematic exploration of Archaea—a major yet underexplored branch of the tree of life—as a source of novel antimicrobial compounds. www.biorxiv.org/content/10.1...
Reposted by Tudor-Stefan Cotet
Excited to finally share Perceptein, a PERCEPtron made of proTEINs: www.science.org/doi/10.1126/...
A synthetic protein-level neural network in mammalian cells
Artificial neural networks provide a powerful paradigm for nonbiological information processing. To understand whether similar principles could enable computation within living cells, we combined de n...
www.science.org
December 13, 2024 at 1:41 PM
Excited to finally share Perceptein, a PERCEPtron made of proTEINs: www.science.org/doi/10.1126/...
Reposted by Tudor-Stefan Cotet
FlowMol at your fingertips! We just released a colab notebook to make using FlowMol super easy. Come chat with us tomorrow at @workshopmlsb ! #NeurIPS2024 🧪 colab.research.google.com/github/Dunni...
December 15, 2024 at 12:34 AM
FlowMol at your fingertips! We just released a colab notebook to make using FlowMol super easy. Come chat with us tomorrow at @workshopmlsb ! #NeurIPS2024 🧪 colab.research.google.com/github/Dunni...
Reposted by Tudor-Stefan Cotet
Adding structure tokens improves zero-shot variant effect prediction of ESM3, but only when structural is experimentally determined and has sequence of interest. Mutant structures from Rosetta or MD (<1 Å RMSD) worsen prediction compared to running w/ only sequence doi.org/10.1101/2024.12.09.627585
December 10, 2024 at 6:25 AM
Adding structure tokens improves zero-shot variant effect prediction of ESM3, but only when structural is experimentally determined and has sequence of interest. Mutant structures from Rosetta or MD (<1 Å RMSD) worsen prediction compared to running w/ only sequence doi.org/10.1101/2024.12.09.627585
Reposted by Tudor-Stefan Cotet
After two years, our paper on generative models for structure-based drug design is finally out in @natcomputsci.bsky.social
www.nature.com/articles/s43...
www.nature.com/articles/s43...
Structure-based drug design with equivariant diffusion models - Nature Computational Science
This work applies diffusion models to conditional molecule generation and shows how they can be used to tackle various structure-based drug design problems
www.nature.com
December 9, 2024 at 2:00 PM
After two years, our paper on generative models for structure-based drug design is finally out in @natcomputsci.bsky.social
www.nature.com/articles/s43...
www.nature.com/articles/s43...
Reposted by Tudor-Stefan Cotet
Another great model from @msftresearch.bsky.social AI for Science - CHIMERA, an accurate retrosynthesis prediction model by @marwinsegler @MaziarzKris and team!
new preprint on chemical synthesis ML models
- showing how to combine multiple models in a principled way
- modern Transformers + GNN to featurize chemical reaction:
- new insights in where the models shine
+ bonus: find the quirky named reaction!
Feedback welcome!
arxiv.org/abs/2412.05269
- showing how to combine multiple models in a principled way
- modern Transformers + GNN to featurize chemical reaction:
- new insights in where the models shine
+ bonus: find the quirky named reaction!
Feedback welcome!
arxiv.org/abs/2412.05269
Chimera: Accurate retrosynthesis prediction by ensembling models with diverse inductive biases
Planning and conducting chemical syntheses remains a major bottleneck in the discovery of functional small molecules, and prevents fully leveraging generative AI for molecular inverse design. While ea...
arxiv.org
December 9, 2024 at 3:11 PM
Another great model from @msftresearch.bsky.social AI for Science - CHIMERA, an accurate retrosynthesis prediction model by @marwinsegler @MaziarzKris and team!
Reposted by Tudor-Stefan Cotet
Hi @ddelalamo.bsky.social unfortunately, this paper from Jain et al. contains falsehoods, misleading comparisons, seemingly deliberate omissions, and is written in a tone not intended as a serious research paper. Please see our detailed response: www.linkedin.com/pulse/respon...
Response to Jain et al.
You may have seen a recent pre-print [1] from Jain et al. with strongly worded claims against the experimental results in our DiffDock paper [2].
www.linkedin.com
December 8, 2024 at 9:46 PM
Hi @ddelalamo.bsky.social unfortunately, this paper from Jain et al. contains falsehoods, misleading comparisons, seemingly deliberate omissions, and is written in a tone not intended as a serious research paper. Please see our detailed response: www.linkedin.com/pulse/respon...
Reposted by Tudor-Stefan Cotet
De novo protein synthesis for specific recognition of peptide-MHC complex! This represents the potential for significant advancement in the development of immunotherapies!
doi: doi.org/10.1101/2024...
doi: doi.org/10.1101/2024...
doi.org
December 9, 2024 at 6:40 AM
De novo protein synthesis for specific recognition of peptide-MHC complex! This represents the potential for significant advancement in the development of immunotherapies!
doi: doi.org/10.1101/2024...
doi: doi.org/10.1101/2024...
Reposted by Tudor-Stefan Cotet
Exciting results from Adaptyv’s second round of the binder design challange! Top ranked binders are optimised antibodies, nanobodies, or native interaction partners! Congratulations to all the winners, well deserved!!! 🚀🚀
foundry.adaptyvbio.com/competition
foundry.adaptyvbio.com/competition
Adaptyv Foundry
foundry.adaptyvbio.com
December 8, 2024 at 7:29 AM
Exciting results from Adaptyv’s second round of the binder design challange! Top ranked binders are optimised antibodies, nanobodies, or native interaction partners! Congratulations to all the winners, well deserved!!! 🚀🚀
foundry.adaptyvbio.com/competition
foundry.adaptyvbio.com/competition
Reposted by Tudor-Stefan Cotet
Give everyone exactly the same problem. Love seeing the diversity in strategy and tools used.
It’s refreshing as most of us are always set on doing things only one way.
You always think back at the end and wonder if you only had tackled it differently from the start www.adaptyvbio.com/blog/po103
It’s refreshing as most of us are always set on doing things only one way.
You always think back at the end and wonder if you only had tackled it differently from the start www.adaptyvbio.com/blog/po103
Adaptyv Bio - Protein Optimization 103: Racing to the Top 100
What a close race this one was! In this blog post, we look at how submissions evolved throughout our latest EGFR binder design competition. We highlight the most widespread model and design choices, h...
www.adaptyvbio.com
December 6, 2024 at 8:27 AM
Give everyone exactly the same problem. Love seeing the diversity in strategy and tools used.
It’s refreshing as most of us are always set on doing things only one way.
You always think back at the end and wonder if you only had tackled it differently from the start www.adaptyvbio.com/blog/po103
It’s refreshing as most of us are always set on doing things only one way.
You always think back at the end and wonder if you only had tackled it differently from the start www.adaptyvbio.com/blog/po103
Reposted by Tudor-Stefan Cotet
📢 New #preprint describing our community effort to share #plumed tutorials arxiv.org/abs/2412.03595 ! Explore the tutorials at www.plumed-tutorials.org
PLUMED Tutorials: a collaborative, community-driven learning ecosystem
In computational physics, chemistry, and biology, the implementation of new techniques in a shared and open source software lowers barriers to entry and promotes rapid scientific progress. However, ef...
arxiv.org
December 6, 2024 at 3:07 PM
📢 New #preprint describing our community effort to share #plumed tutorials arxiv.org/abs/2412.03595 ! Explore the tutorials at www.plumed-tutorials.org
Reposted by Tudor-Stefan Cotet
1/🧬 Excited to share PLAID, our new approach for co-generating sequence and all-atom protein structures by sampling from the latent space of ESMFold. This requires only sequences during training, which unlocks more data and annotations:
bit.ly/plaid-proteins
🧵
bit.ly/plaid-proteins
🧵
December 6, 2024 at 5:44 PM
1/🧬 Excited to share PLAID, our new approach for co-generating sequence and all-atom protein structures by sampling from the latent space of ESMFold. This requires only sequences during training, which unlocks more data and annotations:
bit.ly/plaid-proteins
🧵
bit.ly/plaid-proteins
🧵
Reposted by Tudor-Stefan Cotet
Great blog post from AdaptyvBio summarising submissions for round 2 of the protein design competition: including a few interesting methods I had never heard of. Results will be released tomorrow:
www.adaptyvbio.com/blog/po103
www.adaptyvbio.com/blog/po103
Adaptyv Bio - Protein Optimization 103: Racing to the Top 100
What a close race this one was! In this blog post, we look at how submissions evolved throughout our latest EGFR binder design competition. We highlight the most widespread model and design choices, h...
www.adaptyvbio.com
December 6, 2024 at 8:53 AM
Great blog post from AdaptyvBio summarising submissions for round 2 of the protein design competition: including a few interesting methods I had never heard of. Results will be released tomorrow:
www.adaptyvbio.com/blog/po103
www.adaptyvbio.com/blog/po103
Reposted by Tudor-Stefan Cotet
Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @msftresearch.bsky.social ch AI for Science.
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
December 6, 2024 at 8:39 AM
Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @msftresearch.bsky.social ch AI for Science.
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
Reposted by Tudor-Stefan Cotet
Excited to present what we've been up to the last couple years. Introducing BioEmu, a Biomolecular Emulator of protein dynamics: www.biorxiv.org/content/10.1...
Scalable emulation of protein equilibrium ensembles with generative deep learning
Following the sequence and structure revolutions, predicting the dynamical mechanisms of proteins that implement biological function remains an outstanding scientific challenge. Several experimental t...
www.biorxiv.org
December 6, 2024 at 8:22 AM
Excited to present what we've been up to the last couple years. Introducing BioEmu, a Biomolecular Emulator of protein dynamics: www.biorxiv.org/content/10.1...
Reposted by Tudor-Stefan Cotet
We are thrilled to share ProDomino a model for the prediction of domain insertion sites in proteins. Our approach enables the simple and rapid engineering of highly potent switchable proteins, as we exemplify by creating novel inducible variants of Cas9 and Cas12a.
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
Rational engineering of allosteric protein switches by in silico prediction of domain insertion sites
Domain insertion engineering is a powerful approach to juxtapose otherwise separate biological functions, resulting in proteins with new-to-nature activities. A prominent example are switchable protei...
www.biorxiv.org
December 5, 2024 at 9:59 AM
We are thrilled to share ProDomino a model for the prediction of domain insertion sites in proteins. Our approach enables the simple and rapid engineering of highly potent switchable proteins, as we exemplify by creating novel inducible variants of Cas9 and Cas12a.
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
Reposted by Tudor-Stefan Cotet
Been a while since I read a paper like this:
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
Deep-Learning Based Docking Methods: Fair Comparisons to Conventional Docking Workflows
The diffusion learning method, DiffDock, for docking small-molecule ligands into protein binding sites was recently introduced. Results included comparisons to more conventional docking approaches, wi...
arxiv.org
December 5, 2024 at 3:36 PM
Been a while since I read a paper like this:
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889
• "What [DiffDock] appears to be doing cannot be considered" docking
• "Results are ... contaminated with near neighbors to test cases"
• "Results for DiffDock were artifactual"
• "Results for other methods were incorrectly done"
arxiv.org/abs/2412.02889