pith. sign in

arxiv: 2606.18703 · v1 · pith:BL5EWQCFnew · submitted 2026-06-17 · 💻 cs.LG · q-bio.QM

Contextualizing Biological Language Models across Modalities via Logit-Space Contrastive Alignment

Yanjun Shao , Yundi Chen , Yashvi Patel , Aurelien Pelissier , Mar\'ia Rodr\'iguez Mart\'inez This is my paper

Pith reviewed 2026-06-26 21:29 UTC · model grok-4.3

classification 💻 cs.LG q-bio.QM
keywords biological language modelslogit-space contrastive alignmentcontextual predictionvariant rankingdrug resistance predictionprotein-ligand bindingTCR-peptide activity
0
0 comments X

The pith

LOGICA aligns biological language models in logit space to add context from ligands or drugs while preserving the original per-token likelihood interface.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces LOGICA to make pretrained biological language models respond to task-specific contexts such as interaction partners or therapeutic interventions. It does so by running contrastive alignment directly on the models' output logits using gated adapters that match each model's native vocabulary and token head. This setup allows training on sparse paired data across modalities without a shared tokenizer or any change to the base likelihood interface. The resulting context-sensitive token probabilities improve performance on variant ranking tasks including protein-ligand binding, TCR-peptide activity, and drug-conditioned resistance prediction. A reader would care because the method keeps the models usable for sequence design and mechanistic interpretation instead of replacing their probability outputs with pooled embeddings or new heads.

Core claim

LOGICA performs contrastive learning directly in output-logit space with gated cross-modal adapters that interface with each model's native token head, producing context-conditioned token probabilities that improve mutation-local variant ranking on protein-ligand, TCR-peptide, and drug-resistance tasks while preserving the pretrained per-token likelihood interface and requiring no shared tokenizer.

What carries the argument

Gated cross-modal adapters that map contextual inputs into adjustments of each base model's logit outputs, enabling contrastive alignment of token probabilities across modalities.

If this is right

  • Mutation-local variant ranking reduces to direct comparison of context-conditioned token likelihoods at the perturbed site.
  • Models with distinct vocabularies can be aligned for joint prediction using only sparse paired examples.
  • The native token-level interface remains available for both mechanistic interpretation and sequence generation.
  • AUC on held-out-gene single-mutation drug-resistance prediction rises from near-random latent baselines of ~0.55 to ~0.65.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same logit-space alignment could be tested on additional modalities such as gene-expression or metabolite contexts to check whether the gains generalize beyond the three tasks studied.
  • Working in probability space rather than embedding space may reduce the data needed to combine separately pretrained biological models.
  • The approach could be applied to score multi-mutation combinations under drug pressure without retraining the base language model.

Load-bearing premise

Gated adapters can align logits across models with different vocabularies using only sparse paired data without distorting the original per-token probability distributions.

What would settle it

On a new held-out set of single-mutation drug-resistance cases, if the context-conditioned token likelihoods produced by LOGICA rank true resistant variants no higher than the uncontextualized base model, the central claim would be falsified.

Figures

Figures reproduced from arXiv: 2606.18703 by Aurelien Pelissier, Mar\'ia Rodr\'iguez Mart\'inez, Yanjun Shao, Yashvi Patel, Yundi Chen.

Figure 1
Figure 1. Figure 1: Overview of LOGICA: pretrained biological language models are coupled by cross-modal adapters that preserve native token heads, enabling contrastive alignment of context-conditioned logit-probability distributions across distinct modal vocabularies. 2 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Two scaling regimes for protein–ligand LOGICA. (A) Held-out likelihood-margin trajec￾tories during pretraining for ESM-2 backbones at {8, 35, 150, 650}M parameters. (B) Peak margin versus backbone size; dashed line shows a log-log fit, γb¯ ⋆ ∝ N0.16. (C) Few-shot drug-resistance ranking as a fraction of target-gene labels is used for adaptation and the remaining variants are held out for evaluation. 3.2 TC… view at source ↗
Figure 3
Figure 3. Figure 3: LOGICA performs zero-shot TCR–peptide variant ranking and identifies cross￾modal dependencies. (A, B) Pairwise win-margin heatmap for peptide variant ranking and TCR variant ranking. Each cell reports the mean difference in Spearman correlation between the row model and the column model across mutation sets, with positive values indicating that the row model performs better. Individual scores are provided … view at source ↗
read the original abstract

Pretrained biological language models expose per-token probability distributions through masked-token prediction, providing the likelihood interface central to sequence design, variant scoring, and mechanistic interpretation. Yet these distributions are learned from broad unlabeled corpora and are not naturally conditioned on task-specific biological contexts such as interaction partners, cellular environments, or therapeutic interventions. Existing contextual matching methods often distort this interface through pooled embeddings, contrastive latent spaces, or task-specific prediction heads. We introduce LOGICA (Logit-space Contrastive Alignment), a framework for context-conditioned prediction that performs contrastive learning directly in output-logit space. Using gated cross-modal adapters compatible with each model's native token head, LOGICA preserves the pretrained likelihood interface and converts contextualized token log-likelihoods into matching scores. Alignment is defined through context-sensitive token probabilities rather than proximity in a shared embedding space, enabling learning from sparse paired data across models with distinct vocabularies, without a shared tokenizer or decoder. LOGICA is particularly effective for mutation-local variant ranking, where comparisons reduce to context-conditioned likelihoods of mutant tokens at perturbed sites. Across protein--ligand binding, TCR--peptide activity, and drug-conditioned resistance prediction, LOGICA improves over prior state-of-the-art methods, including matched latent-contrastive and conditional MLM baselines, while retaining a token-level interface for interpretation and generation. On held-out-gene single-mutation drug-resistance prediction, LOGICA improves AUC from near-random latent-space baselines of $\sim$0.55 to $\sim$0.65.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces LOGICA, a logit-space contrastive alignment framework that uses gated cross-modal adapters to condition pretrained biological language models on task-specific contexts (e.g., ligands, peptides, drugs) while preserving the native per-token likelihood interface. It reports AUC gains over latent-space and conditional-MLM baselines on protein-ligand binding, TCR-peptide activity, and held-out-gene single-mutation drug-resistance prediction (0.55 to 0.65), with the alignment performed directly on context-sensitive token probabilities rather than embeddings and without requiring a shared tokenizer.

Significance. If the preservation of the original per-token likelihood surface is empirically validated, the method would provide a practical route to contextualized variant scoring and generation that retains interpretability advantages of the pretrained token heads. The held-out-gene evaluation and cross-vocabulary compatibility are positive design choices that strengthen the generalization claim.

major comments (2)
  1. [Abstract] Abstract: the central claim that gated adapters 'preserve the pretrained likelihood interface' and 'convert contextualized token log-likelihoods into matching scores' without distortion is load-bearing for the interpretation/generation benefit, yet no direct metric (KL divergence, rank correlation, or calibration error) is reported comparing pre- and post-adapter per-token log-probabilities on held-out unpaired sequences outside the paired training distribution.
  2. [Abstract] Abstract: the reported AUC improvement (∼0.55 to ∼0.65) on held-out-gene drug-resistance prediction lacks accompanying details on training-set sizes, number of paired examples, statistical significance, error bars, or ablation controls on the adapter architecture, making it impossible to assess whether the gain arises from logit-space alignment or from other factors.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'near-random latent-space baselines of ∼0.55' should be replaced by the exact baseline values and the precise latent-space method used for comparison.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that gated adapters 'preserve the pretrained likelihood interface' and 'convert contextualized token log-likelihoods into matching scores' without distortion is load-bearing for the interpretation/generation benefit, yet no direct metric (KL divergence, rank correlation, or calibration error) is reported comparing pre- and post-adapter per-token log-probabilities on held-out unpaired sequences outside the paired training distribution.

    Authors: We agree that direct quantitative validation of likelihood preservation on held-out unpaired sequences is a valuable addition. The gated adapter design inserts a residual connection from the original logits, which by construction leaves the token head unchanged, but we acknowledge the absence of explicit metrics such as KL divergence or rank correlation in the current version. In revision we will add these metrics (KL, Spearman correlation of token ranks, and expected calibration error) computed on held-out sequences drawn from the pretraining distribution, reported in a new results subsection and referenced from the abstract. revision: yes

  2. Referee: [Abstract] Abstract: the reported AUC improvement (∼0.55 to ∼0.65) on held-out-gene drug-resistance prediction lacks accompanying details on training-set sizes, number of paired examples, statistical significance, error bars, or ablation controls on the adapter architecture, making it impossible to assess whether the gain arises from logit-space alignment or from other factors.

    Authors: We will revise the abstract to incorporate the requested details: approximate number of paired examples for the drug-resistance task, error bars from repeated runs, and a statement of statistical significance. Ablation results on adapter components (gating, contrastive objective) are already present in the supplementary material; we will add an explicit cross-reference in the abstract. These changes will allow readers to evaluate the source of the observed improvement without altering the core claims. revision: yes

Circularity Check

0 steps flagged

No circularity; new training procedure with independent empirical results

full rationale

The paper introduces LOGICA as a novel contrastive alignment framework operating directly in output-logit space via gated adapters. No equations, derivations, or self-citations in the provided text reduce the reported AUC gains (0.55 to 0.65) or the preservation of the token-level interface to quantities defined by construction from the same fitted parameters or prior self-referential results. The method is presented as an independent training procedure whose value is assessed via downstream task performance on held-out data, without any load-bearing step that renames a fit as a prediction or imports uniqueness via author-overlapping citations. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are stated. The central claim rests on the unstated assumption that logit-space contrastive alignment preserves native likelihoods across modalities.

pith-pipeline@v0.9.1-grok · 5823 in / 1290 out tokens · 19057 ms · 2026-06-26T21:29:08.784122+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

99 extracted references · 90 canonical work pages · 5 internal anchors

  1. [1]

    Evolutionary-scale prediction of atomic-level protein structure with a language model,

    Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Nikita Smetanin, Robert Verkuil, Ori Kabeli, Yaniv Shmueli, Allan Dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Salvatore Candido, and Alexander Rives. Evolutionary-scale prediction of atomic- level protein structure with a language model.Science, 379(6637):1123–1130, 2023. doi:...

    work page doi:10.1126/science.ade2574 2023

  2. [2]

    Nature Methods , volume =

    Adam J Riesselman, John B Ingraham, and Debora S Marks. Deep generative models of genetic variation capture the effects of mutations.Nature methods, 15(10):816–822, 2018. doi: 10.1038/s41592-018-0138-4. URLhttps://doi.org/10.1038/s41592-018-0138-4

    work page doi:10.1038/s41592-018-0138-4 2018

  3. [3]

    2021 , publisher =

    Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu, and Alexander Rives. Lan- guage models enable zero-shot prediction of the effects of mutations on protein function. In Advances in Neural Information Processing Systems, volume 34, pages 29287–29303, 2021. doi: 10.1101/2021.07.09.450648. URLhttps://doi.org/10.1101/2021.07.09.450648

    work page doi:10.1101/2021.07.09.450648 2021

  4. [4]

    Hie, Varun R

    Brian L. Hie, Varun R. Shanker, Duo Xu, Theodora U. J. Bruun, Payton A. Weiden- bacher, Shaogeng Tang, Wesley Wu, John E. Pak, and Peter S. Kim. Efficient evolution of human antibodies from general protein language models.Nature Biotechnology, 42(2): 275–283, 2024. doi: 10.1038/s41587-023-01763-2. URLhttps://doi.org/10.1038/ s41587-023-01763-2

    work page doi:10.1038/s41587-023-01763-2 2024

  5. [5]

    Hie, Kevin K

    Brian L. Hie, Kevin K. Yang, and Peter S. Kim. Evolutionary velocity with protein language models predicts evolutionary dynamics of diverse proteins.Cell Systems, 13(4):274–285,

  6. [6]

    URLhttps://doi.org/10.1016/j.cels.2022

    doi: 10.1016/j.cels.2022.01.003. URLhttps://doi.org/10.1016/j.cels.2022. 01.003

    work page doi:10.1016/j.cels.2022.01.003 2022

  7. [7]

    SELFormer: Molecular rep- resentation learning via SELFIES language models.Machine Learning: Science and Tech- nology, 4(2):025035, 2023

    Atakan Y ¨uksel, Erva Ulusoy, Atabey ¨Unl¨u, and Tunca Do ˘gan. SELFormer: Molecular rep- resentation learning via SELFIES language models.Machine Learning: Science and Tech- nology, 4(2):025035, 2023. doi: 10.1088/2632-2153/acdb30. URLhttps://doi.org/10. 1088/2632-2153/acdb30

    work page doi:10.1088/2632-2153/acdb30 2023

  8. [8]

    de Almeida, Hassan Sirelkhatim, Guillaume Richard, Marcin Skwark, Karim Beguir, Marie Lopez, and Thomas Pierrot

    Hugo Dalla-Torre, Liam Gonzalez, Javier Mendoza-Revilla, Nicolas Lopez Carranza, Adam Henryk Grzywaczewski, Francesco Oteri, Christian Dallago, Evan Trop, Bernardo P. de Almeida, Hassan Sirelkhatim, Guillaume Richard, Marcin Skwark, Karim Beguir, Marie Lopez, and Thomas Pierrot. Nucleotide transformer: building and evaluating robust foun- dation models fo...

    work page doi:10.1038/s41592-024-02523-z 2025

  9. [9]

    Claire Donnat and Elena Tuzhilina

    Haotian Cui, Alejandro Tejada-Lapuerta, Maria Brbi ´c, Julio Saez-Rodriguez, Simona Cristea, Hani Goodarzi, Mohammad Lotfollahi, Fabian J Theis, and Bo Wang. Towards multimodal foundation models in molecular cell biology.Nature, 640(8059):623–633, 2025. doi: 10.1038/ s41586-025-08710-y. URLhttps://doi.org/10.1038/s41586-025-08710-y

    work page doi:10.1038/s41586-025-08710-y 2025

  10. [10]

    Durrant, Jerome Ku, Mohsen Naghipourfar, Michael Poli, Gwang- gyu Sun, Greg Brockman, Daniel Chang, Alison Fanton, Gabriel A

    Garyk Brixi, Matthew G. Durrant, Jerome Ku, Mohsen Naghipourfar, Michael Poli, Gwang- gyu Sun, Greg Brockman, Daniel Chang, Alison Fanton, Gabriel A. Gonzalez, Samuel H. King, David B. Li, Aditi T. Merchant, Eric Nguyen, Chiara Ricci-Tam, David W. Romero, Jonathan C. Schmok, Ali Taghibakhshi, Anton V orontsov, Brandon Yang, Myra Deng, Liv Gorton, Nam Nguy...

    work page doi:10.1038/s41586-026-10176-5 2026

  11. [11]

    Nucleotide dependency analysis of genomic language models detects functional elements.Nature Genetics, 57: 2589–2602, 2025

    Pedro Tomaz da Silva, Alexander Karollus, Johannes Hingerl, Gihanna Galindez, Nils Wag- ner, Xavier Hernandez-Alias, Danny Incarnato, and Julien Gagneur. Nucleotide dependency analysis of genomic language models detects functional elements.Nature Genetics, 57: 2589–2602, 2025. doi: 10.1038/s41588-025-02347-3. URLhttps://doi.org/10.1038/ s41588-025-02347-3

    work page doi:10.1038/s41588-025-02347-3 2025

  12. [12]

    Klivans, James Madigan Loy, Tianlong Chen, Qiang Liu, and Daniel Jesus Diaz

    Chengyue Gong, Adam R. Klivans, James Madigan Loy, Tianlong Chen, Qiang Liu, and Daniel Jesus Diaz. Evolution-inspired loss functions for protein representation learning. In Proceedings of the 41st International Conference on Machine Learning (ICML), volume 235 ofProceedings of Machine Learning Research, pages 15893–15906. PMLR, 2024. URL https://proceedi...

    2024

  13. [13]

    Wayment-Steele, Garyk Brixi, Hong Wang, David Kern, and Sergey Ovchinnikov

    Zhidian Zhang, Hannah K. Wayment-Steele, Garyk Brixi, Hong Wang, David Kern, and Sergey Ovchinnikov. Protein language models learn evolutionary statistics of interacting se- quence motifs.Proceedings of the National Academy of Sciences, 121(45):e2406285121, 2024. doi: 10.1073/pnas.2406285121. URLhttps://doi.org/10.1073/pnas.2406285121

    work page doi:10.1073/pnas.2406285121 2024

  14. [14]

    Greene, Subu Subramanian, Benjamin P

    Ali Madani, Ben Krause, Eric R. Greene, Subu Subramanian, Benjamin P. Mohr, James M. Holton, Jose Luis Olmos Jr, Caiming Xiong, Zachary Z. Sun, Richard Socher, James S. Fraser, and Nikhil Naik. Large language models generate functional protein sequences across diverse families.Nature biotechnology, 41(8):1099–1106, 2023. doi: 10.1038/s41587-022-01618-2. U...

    work page doi:10.1038/s41587-022-01618-2 2023

  15. [15]

    Generating novel protein sequences using gibbs sampling of masked language models.bioRxiv, pages 2021–01, 2021

    Sean R Johnson, Sarah Monaco, Kenneth Massie, and Zaid Syed. Generating novel protein sequences using gibbs sampling of masked language models.bioRxiv, pages 2021–01, 2021. doi: 10.1101/2021.01.26.428322. URLhttps://doi.org/10.1101/2021.01.26.428322

    work page doi:10.1101/2021.01.26.428322 2021

  16. [16]

    How to make the most of your masked language model for protein engineering

    Calvin McCarter, Nick Bhattacharya, Sebastian W Ober, and Hunter Elliott. How to make the most of your masked language model for protein engineering.arXiv preprint arXiv:2603.10302, 2026. doi: 10.48550/arXiv.2603.10302. URLhttps://arxiv.org/abs/ 2603.10302

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2603.10302 2026

  17. [17]

    Nature , volume =

    Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Boden- stein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasu- vunakool, Zachary Wu, Akvil ˙e ˇZemgulyt˙e, Eirini Arvaniti, Charles Beattie, Ottavia Bertol...

    work page doi:10.1038/s41586-024-07487-w 2024

  18. [18]

    Pan-peptide meta learning for T-cell receptor–antigen binding recognition.Nature Machine Intelligence, 2023

    Yicheng Gao, Yuli Gao, Yuxiao Fan, Chengyu Zhu, Zhiting Wei, Chi Zhou, Guohui Chuai, Qinchang Chen, He Zhang, and Qi Liu. Pan-peptide meta learning for T-cell receptor–antigen binding recognition.Nature Machine Intelligence, 2023. doi: 10.1038/s42256-023-00619-3. URLhttps://doi.org/10.1038/s42256-023-00619-3

    work page doi:10.1038/s42256-023-00619-3 2023

  19. [19]

    Boltz-2: Towards accurate and efficient binding affinity prediction.bioRxiv, 2025

    Saro Passaro, Gabriele Corso, Jeremy Wohlwend, Mateo Reveiz, Stephan Thaler, Vignesh Ram Somnath, Noah Getz, Tally Portnoi, Julien Roy, Hannes Stark, David Kwabi-Addo, Dominique Beaini, Tommi Jaakkola, and Regina Barzilay. Boltz-2: Towards accurate and efficient binding affinity prediction, 2025. URLhttps://doi.org/10.1101/2025.06.14.659707. bioRxiv preprint

    work page doi:10.1101/2025.06.14.659707 2025

  20. [20]

    Glass, and Jimeng Sun

    Kexin Huang, Cao Xiao, Lucas M. Glass, and Jimeng Sun. MolTrans: Molecular in- teraction transformer for drug–target interaction prediction.Bioinformatics, 37(6):830– 836, 2021. doi: 10.1093/bioinformatics/btaa880. URLhttps://doi.org/10.1093/ bioinformatics/btaa880. 11

    work page doi:10.1093/bioinformatics/btaa880 2021

  21. [21]

    Deep contrastive learning enables genome-wide virtual screening.Science, 391(6781):eads9530, 2026

    Yinjun Jia, Bowen Gao, Jiaxin Tan, Jiqing Zheng, Xin Hong, Wenyu Zhu, Haichuan Tan, Yuan Xiao, Liping Tan, Hongyi Cai, Yanwen Huang, Zhiheng Deng, Xiangwei Wu, Yue Jin, Yafei Yuan, Jiekang Tian, Wei He, Weiying Ma, Yaqin Zhang, Lei Liu, Chuangye Yan, Wei Zhang, and Yanyan Lan. Deep contrastive learning enables genome-wide virtual screening.Science, 391(67...

    work page doi:10.1126/science.ads9530 2026

  22. [22]

    Con- trastive learning in protein language space predicts interactions between drugs and protein targets.Proceedings of the National Academy of Sciences, 120(24):e2220778120, 2023

    Rohit Singh, Samuel Sledzieski, Bryan Bryson, Lenore Cowen, and Bonnie Berger. Con- trastive learning in protein language space predicts interactions between drugs and protein targets.Proceedings of the National Academy of Sciences, 120(24):e2220778120, 2023. doi: 10.1073/pnas.2220778120. URLhttps://doi.org/10.1073/pnas.2220778120

    work page doi:10.1073/pnas.2220778120 2023

  23. [23]

    Interpretable bilinear attention net- work with domain adaptation improves drug–target prediction.Nature Machine Intelligence, 5 (2):126–136, 2023

    Peizhen Bai, Filip Miljkovi ´c, Bino John, and Haiping Lu. Interpretable bilinear attention net- work with domain adaptation improves drug–target prediction.Nature Machine Intelligence, 5 (2):126–136, 2023. doi: 10.1038/s42256-022-00605-1. URLhttps://doi.org/10.1038/ s42256-022-00605-1

    work page doi:10.1038/s42256-022-00605-1 2023

  24. [24]

    Sizhe Liu, Yuchen Liu, Haofeng Xu, Jun Xia, and Stan Z. Li. SP-DTI: subpocket-informed transformer for drug–target interaction prediction.Bioinformatics, 41(3):btaf011, 03 2025. ISSN 1367-4811. doi: 10.1093/bioinformatics/btaf011. URLhttps://doi.org/10.1093/ bioinformatics/btaf011

    work page doi:10.1093/bioinformatics/btaf011 2025

  25. [25]

    GS-DTI: A graph-structure-aware framework leveraging large language models for drug–target interaction prediction.Bioinfor- matics, 41(8):btaf445, 08 2025

    Qinze Yu, Chang Zhou, Jiyue Jiang, Xiangyu Shi, and Yu Li. GS-DTI: A graph-structure-aware framework leveraging large language models for drug–target interaction prediction.Bioinfor- matics, 41(8):btaf445, 08 2025. ISSN 1367-4811. doi: 10.1093/bioinformatics/btaf445. URL https://doi.org/10.1093/bioinformatics/btaf445

    work page doi:10.1093/bioinformatics/btaf445 2025

  26. [26]

    Weber, Ella Barkan, Simona Rabinovici-Cohen, Sagi Polaczek, Ido Amos, et al

    Yoel Shoshan, Moshiko Raboh, Michal Ozery-Flato, Vadim Ratner, Alex Golts, Jeffrey K. Weber, Ella Barkan, Simona Rabinovici-Cohen, Sagi Polaczek, Ido Amos, et al. MAMMAL – molecular aligned multi-modal architecture and language for biomedical discovery.npj Drug Discovery, 2026. doi: 10.1038/s44386-026-00047-4. URLhttps://doi.org/10.1038/ s44386-026-00047-4

    work page doi:10.1038/s44386-026-00047-4 2026

  27. [27]

    & Berger, B

    Varun Ullanat, Bowen Jing, Samuel Sledzieski, and Bonnie Berger. Learning the language of protein-protein interactions.Nature Communications, 17:1199, 2026. doi: 10.1038/ s41467-025-67971-3. URLhttps://doi.org/10.1038/s41467-025-67971-3

    work page doi:10.1038/s41467-025-67971-3 2026

  28. [28]

    4M: Massively multimodal masked modeling

    David Mizrahi, Roman Bachmann, O ˘guzhan Fatih Kar, Teresa Yeo, Mingfei Gao, Afshin De- hghan, and Amir Zamir. 4M: Massively multimodal masked modeling. InAdvances in Neu- ral Information Processing Systems (NeurIPS), volume 36, pages 58363–58408, 2023. URL https://arxiv.org/abs/2312.06647

    arXiv 2023

  29. [29]

    Walczak, and Thierry Mora

    Barthelemy Meynard-Piganeau, Christoph Feinauer, Martin Weigt, Aleksandra M. Walczak, and Thierry Mora. TULIP: A transformer-based unsupervised language model for interacting peptides and T cell receptors that generalizes to unseen epitopes.Proceedings of the National Academy of Sciences, 121(24):e2316401121, 2024. doi: 10.1073/pnas.2316401121. URL https:...

    work page doi:10.1073/pnas.2316401121 2024

  30. [30]

    Bennett, Amy G

    Dhuvarakesh Karthikeyan, Sarah N. Bennett, Amy G. Reynolds, Benjamin G. Vincent, and Alex Rubinsteyn. Conditional generation of real antigen-specific T cell receptor sequences. Nature Machine Intelligence, 7(9):1494–1509, 2025. doi: 10.1038/s42256-025-01096-6. URL https://doi.org/10.1038/s42256-025-01096-6

    work page doi:10.1038/s42256-025-01096-6 2025

  31. [31]

    DeLisa, Jen-Tsan Ashley Chi, Ray Truant, Hector C

    Leo Tianlai Chen, Zachary Quinn, Madeleine Dumas, Christina Peng, Lauren Hong, Moi- ses Lopez-Gonzalez, Alexander Mestre, Rio Watson, Sophia Vincoff, Lin Zhao, Jianli Wu, Audrey Stavrand, Mayumi Schaepers-Cheu, Tian Zi Wang, Divya Srijay, Connor Monticello, Pranay Vure, Rishab Pulugurta, Sarah Pertsemlidis, Kseniia Kholina, Shrey Goel, Matthew P. DeLisa, ...

    work page doi:10.1038/s41587-025-02761-2 2025

  32. [32]

    Burbach and Bryan Briney

    Sarah M. Burbach and Bryan Briney. Improving antibody language models with native pairing. Patterns, 5(5):100967, 2024. doi: 10.1016/j.patter.2024.100967. URLhttps://doi.org/ 10.1016/j.patter.2024.100967

    work page doi:10.1016/j.patter.2024.100967 2024

  33. [33]

    Lamb, Adalberto Claudio Quiros, Alexandrina Pancheva, Crispin J

    Dan Liu, Francesca Young, Kieran D. Lamb, Adalberto Claudio Quiros, Alexandrina Pancheva, Crispin J. Miller, Craig Macdonald, David L. Robertson, and Ke Yuan. PLM- interact: extending protein language models to predict protein-protein interactions.Nature Communications, 16(1):9012, 2025. doi: 10.1038/s41467-025-64512-w. URLhttps: //doi.org/10.1038/s41467-...

    work page doi:10.1038/s41467-025-64512-w 2025

  34. [34]

    Pairing interacting pro- tein sequences using masked language modeling.Proceedings of the National Academy of Sciences, 121(27):e2311887121, 2024

    Umberto Lupo, Damiano Sgarbossa, and Anne-Florence Bitbol. Pairing interacting pro- tein sequences using masked language modeling.Proceedings of the National Academy of Sciences, 121(27):e2311887121, 2024. doi: 10.1073/pnas.2311887121. URLhttps: //doi.org/10.1073/pnas.2311887121

    work page doi:10.1073/pnas.2311887121 2024

  35. [35]

    Matthew I. J. Raybould, Alexander Greenshields-Watson, Parth Agarwal, Broncio Aguilar- Sanjuan, Tobias H. Olsen, Oliver M. Turnbull, Nele P. Quast, and Charlotte M. Deane. The observed T cell receptor space database enables paired-chain repertoire mining, coherence analysis, and language modeling.Cell Reports, 43(9):114704, 2024. doi: 10.1016/j.celrep. 20...

    work page doi:10.1016/j.celrep 2024

  36. [36]

    Representation Learning with Contrastive Predictive Coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Representation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748, 2018. doi: 10.48550/arXiv.1807.03748. URLhttps://arxiv.org/abs/1807.03748

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.03748 2018

  37. [37]

    Ralph Allan Bradley and Milton E. Terry. Rank analysis of incomplete block designs: I. The method of paired comparisons.Biometrika, 39(3/4):324–345, 1952. doi: 10.2307/2334029. URLhttps://doi.org/10.2307/2334029

    work page doi:10.2307/2334029 1952

  38. [38]

    Learning to rank using gradient descent

    Chris Burges, Tal Shaked, Erin Renshaw, Ari Lazier, Matt Deeds, Nicole Hamilton, and Greg Hullender. Learning to rank using gradient descent. InProceedings of the 22nd International Conference on Machine Learning (ICML), pages 89–96. Association for Computing Machin- ery, 2005. doi: 10.1145/1102351.1102363. URLhttps://doi.org/10.1145/1102351. 1102363

    work page doi:10.1145/1102351.1102363 2005

  39. [39]

    Direct Preference Optimization: Your Language Model is Secretly a Reward Model

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D. Manning, Stefano Ermon, and Chelsea Finn. Direct preference optimization: Your language model is secretly a reward model. InAdvances in Neural Information Processing Systems, volume 36, pages 53728–53741, 2023. doi: 10.52202/075280-2338. URLhttps://arxiv.org/abs/2305.18290

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.52202/075280-2338 2023

  40. [40]

    Learning Transferable Visual Models From Natural Language Supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agar- wal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In Proceedings of the 38th International Conference on Machine Learning (ICML), volume 139 ofProce...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2103.00020 2021

  41. [41]

    Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing,

    Hao Tan and Mohit Bansal. Lxmert: Learning cross-modality encoder representations from transformers. InProceedings of the 2019 conference on empirical methods in natural lan- guage processing and the 9th international joint conference on natural language process- ing (EMNLP-IJCNLP), pages 5100–5111, 2019. doi: 10.18653/v1/D19-1514. URLhttps: //arxiv.org/a...

    work page doi:10.18653/v1/d19-1514 2019

  42. [42]

    de Almeida, Lorenz Hexemer, Christopher Blum, Stefan Laurent, Jan Grzegorzewski, Maren Lang, Thomas Pierrot, and Guillaume Richard

    Juan Jose Garau-Luis, Patrick Bordes, Liam Gonzalez, Masa Roller, Bernardo P. de Almeida, Lorenz Hexemer, Christopher Blum, Stefan Laurent, Jan Grzegorzewski, Maren Lang, Thomas Pierrot, and Guillaume Richard. Multi-modal transfer learning between biological foundation models.Advances in Neural Information Processing Systems, 37:78431–78450, 2024. doi: 10...

    work page doi:10.48550/arxiv.2406.14150 2024

  43. [43]

    ChemBERTa: Large-scale self- supervised pretraining for molecular property prediction.arXiv preprint arXiv:2010.09885,

    Seyone Chithrananda, Gabriel Grand, and Bharath Ramsundar. ChemBERTa: Large-scale self- supervised pretraining for molecular property prediction.arXiv preprint arXiv:2010.09885,

    arXiv 2010

  44. [44]

    Chemberta: Large-scale self-supervised pretraining for molecular property prediction.arXiv, 2010.09885, 2020

    doi: 10.48550/arXiv.2010.09885. URLhttps://arxiv.org/abs/2010.09885. 13

    work page doi:10.48550/arxiv.2010.09885 2010

  45. [45]

    Self-referencing embedded strings (SELFIES): A 100% robust molecular string representation

    Mario Krenn, Florian H ¨ase, AkshatKumar Nigam, Pascal Friederich, and Alan Aspuru-Guzik. Self-referencing embedded strings (SELFIES): A 100% robust molecular string representation. Machine Learning: Science and Technology, 1(4):045024, 2020. doi: 10.1088/2632-2153/ aba947. URLhttps://doi.org/10.1088/2632-2153/aba947

    work page doi:10.1088/2632-2153/ 2020

  46. [46]

    Jorissen, and Michael K

    Tiqing Liu, Yuhmei Lin, Xin Wen, Robert N. Jorissen, and Michael K. Gilson. Bind- ingDB in 2024: a FAIR knowledgebase of protein-small molecule binding data.Nucleic Acids Research, 53(D1):D1633–D1644, 2025. doi: 10.1093/nar/gkae1075. URLhttps: //doi.org/10.1093/nar/gkae1075

    work page doi:10.1093/nar/gkae1075 2024

  47. [47]

    Davis, Jeremy P

    Mindy I. Davis, Jeremy P. Hunt, Soren Herrgard, Pietro Ciceri, Lisa M. Wodicka, Gabriel Pallares, Michael Hocker, Daniel K. Treiber, and Patrick P. Zarrinkar. Comprehensive analysis of kinase inhibitor selectivity.Nature Biotechnology, 29(11):1046–1051, 2011. doi: 10.1038/ nbt.1990. URLhttps://doi.org/10.1038/nbt.1990

    work page doi:10.1038/nbt.1990 2011

  48. [48]

    BioSNAP datasets: Stanford biomedical network dataset collection.https://snap.stanford.edu/biodata/,

    Marinka Zitnik, Rok Sosic, Sagar Maheshwari, and Jure Leskovec. BioSNAP datasets: Stanford biomedical network dataset collection.https://snap.stanford.edu/biodata/,

  49. [49]

    URLhttps://snap.stanford.edu/biodata/

  50. [50]

    Coelho, Magdalena E

    Matthew A. Coelho, Magdalena E. Strauss, Alex Watterson, Sarah Cooper, Shriram Bhosle, Giuditta Illuzzi, Emre Karakoc, Cansu Dinc ¸er, Sara F. Vieira, Mamta Sharma, Marie Moullet, Daniela Conticelli, Jonas Koeppel, Katrina McCarten, Chiara M. Cattaneo, Vivien Veninga, Gabriele Picco, Leopold Parts, Josep V . Forment, Emile E. V oest, John C. Marioni, An- ...

    work page doi:10.1038/s41588-024-01948-8 2024

  51. [51]

    Saturation profiling of drug-resistant genetic variants using prime editing.Nature Biotechnology, 43(9): 1471–1484, 2025

    Younggwang Kim, Hyeong-Cheol Oh, Seungho Lee, and Hyongbum Henry Kim. Saturation profiling of drug-resistant genetic variants using prime editing.Nature Biotechnology, 43(9): 1471–1484, 2025. doi: 10.1038/s41587-024-02465-z. URLhttps://doi.org/10.1038/ s41587-024-02465-z

    work page doi:10.1038/s41587-024-02465-z 2025

  52. [52]

    Nature , volume =

    Jonathan Frazer, Pascal Notin, Mafalda Dias, Aidan Gomez, Joseph K. Min, Kevin Brock, Yarin Gal, and Debora S. Marks. Disease variant prediction with deep generative models of evolutionary data.Nature, 599(7883):91–95, 2021. doi: 10.1038/s41586-021-04043-8. URL https://doi.org/10.1038/s41586-021-04043-8

    work page doi:10.1038/s41586-021-04043-8 2021

  53. [53]

    Gomez, Debora Marks, and Yarin Gal

    Pascal Notin, Mafalda Dias, Jonathan Frazer, Javier Marchena-Hurtado, Aidan N. Gomez, Debora Marks, and Yarin Gal. Tranception: Protein fitness prediction with autoregressive transformers and inference-time retrieval. InProceedings of the 39th International Conference on Machine Learning (ICML), volume 162 ofProceedings of Machine Learning Research, pages...

    work page doi:10.48550/arxiv.2205.13760 2022

  54. [54]

    Pattinson, Cornelia L

    Amitava Banerjee, David J. Pattinson, Cornelia L. Wincek, Paul Bunk, Armend Axhemi, Sarah R. Chapin, Saket Navlakha, and Hannah V . Meyer. T cell receptor cross-reactivity pre- diction improved by a comprehensive mutational scan database.Cell Systems, 16(8):101345,

  55. [55]

    URLhttps://doi.org/10.1016/j.cels.2025

    doi: 10.1016/j.cels.2025.101345. URLhttps://doi.org/10.1016/j.cels.2025. 101345

    work page doi:10.1016/j.cels.2025.101345 2025

  56. [56]

    Overton, Sandeep Kumar Dhanda, Sheridan Martini, Jason R

    Randi Vita, Swapnil Mahajan, James A. Overton, Sandeep Kumar Dhanda, Sheridan Martini, Jason R. Cantrell, Daniel K. Wheeler, Alessandro Sette, and Bjoern Peters. The Immune Epitope Database (IEDB): 2018 update.Nucleic Acids Research, 47(D1):D339–D343, 2019. doi: 10.1093/nar/gky1006. URLhttps://doi.org/10.1093/nar/gky1006

    work page doi:10.1093/nar/gky1006 2018

  57. [57]

    Littler, Mark Gerstein, Anthony W

    Yumeng Zhang, Zhikang Wang, Yunzhe Jiang, Dene R. Littler, Mark Gerstein, Anthony W. Purcell, Jamie Rossjohn, Hong-Yu Ou, and Jiangning Song. Epitope-anchored contrastive transfer learning for paired CD8+ T cell receptor–antigen recognition.Nature Machine Intelligence, 6(11):1344–1358, 2024. doi: 10.1038/s42256-024-00913-8. URLhttps: //doi.org/10.1038/s42...

    work page doi:10.1038/s42256-024-00913-8 2024

  58. [58]

    Bjørn P. Y . Kwee, Marius Messemaker, Eric Marcus, Giacomo Oliveira, Wouter Scheper, Catherine J. Wu, Jonas Teuwen, and Ton N. Schumacher. STAPLER: Efficient learning of TCR-peptide specificity prediction from full-length TCR-peptide data.bioRxiv, page 2023.04.25.538237, 2023. doi: 10.1101/2023.04.25.538237. URLhttps://doi.org/10. 1101/2023.04.25.538237

    work page doi:10.1101/2023.04.25.538237 2023

  59. [59]

    The pitfalls of negative data bias for the T-cell epitope specificity challenge.Nature Machine Intelligence, 2023

    Ceder Dens, Kris Laukens, Wout Bittremieux, and Pieter Meysman. The pitfalls of negative data bias for the T-cell epitope specificity challenge.Nature Machine Intelligence, 2023. doi: 10.1038/s42256-023-00727-0. URLhttps://doi.org/10.1038/s42256-023-00727-0

    work page doi:10.1038/s42256-023-00727-0 2023

  60. [60]

    Altin, Coos A

    Eve Richardson, Yannick Jurriaan Maria Aarts, John A. Altin, Coos A. B. Baakman, Philip Bradley, Binbin Chen, Joakim Clifford, Manjima Dhar, Danielle Diepenbroek, Ethan Fast, Ragul Gowthaman, Jieling He, Vadim Karnaukhov, Dario F. Marzella, Pieter Meysman, Morten Nielsen, Jonas Birkelund Nilsson, Sebastian Nymann Deleuran, Farzaneh M. Parizi, Aurelien Pel...

    work page doi:10.64898/2026.03.30.715276 2026

  61. [61]

    Benchmarking of T cell receptor–epitope predictors with ePytope-TCR

    Felix Drost, Anna Chernysheva, Mahmoud Albahah, Katharina Kocher, Kilian Schober, and Benjamin Schubert. Benchmarking of T cell receptor–epitope predictors with ePytope-TCR. Cell Genomics, 5(8):100946, 2025. doi: 10.1016/j.xgen.2025.100946. URLhttps://doi. org/10.1016/j.xgen.2025.100946

    work page doi:10.1016/j.xgen.2025.100946 2025

  62. [62]

    Pierce, Brian M

    Tyler Borrman, Jennifer Cimons, Michael Cosiano, Michael Purcaro, Brian G. Pierce, Brian M. Baker, and Zhiping Weng. ATLAS: a database linking binding affinities with structures for wild-type and mutant TCR–pMHC complexes.Proteins: Structure, Function, and Bioinfor- matics, 85(5):908–916, 2017. doi: 10.1002/prot.25260. URLhttps://doi.org/10.1002/ prot.25260

    work page doi:10.1002/prot.25260 2017

  63. [63]

    Contribution of T cell receptor alpha and beta CDR3, MHC typing, V and J genes to peptide binding prediction.Frontiers in Im- munology, 12:664514, 2021

    Ido Springer, Nili Tickotsky, and Yoram Louzoun. Contribution of T cell receptor alpha and beta CDR3, MHC typing, V and J genes to peptide binding prediction.Frontiers in Im- munology, 12:664514, 2021. doi: 10.3389/fimmu.2021.664514. URLhttps://doi.org/ 10.3389/fimmu.2021.664514

    work page doi:10.3389/fimmu.2021.664514 2021

  64. [64]

    Current challenges for unseen-epitope TCR interaction prediction and a new perspective derived from image clas- sification.Briefings in Bioinformatics, 22(4):bbaa318, 2021

    Pieter Moris, Joey De Pauw, Anna Postovskaya, Sofie Gielis, Nicolas De Neuter, Wout Bit- tremieux, Benson Ogunjimi, Kris Laukens, and Pieter Meysman. Current challenges for unseen-epitope TCR interaction prediction and a new perspective derived from image clas- sification.Briefings in Bioinformatics, 22(4):bbaa318, 2021. doi: 10.1093/bib/bbaa318. URL http...

    work page doi:10.1093/bib/bbaa318 2021

  65. [65]

    iTCep: a deep learning framework for identification of T cell epitopes by harnessing fusion features.Frontiers in Genetics, 14:1141535, 2023

    Yu Zhang, Xingxing Jian, Linfeng Xu, Jingjing Zhao, Manman Lu, Yong Lin, and Lu Xie. iTCep: a deep learning framework for identification of T cell epitopes by harnessing fusion features.Frontiers in Genetics, 14:1141535, 2023. doi: 10.3389/fgene.2023.1141535. URL https://doi.org/10.3389/fgene.2023.1141535

    work page doi:10.3389/fgene.2023.1141535 2023

  66. [66]

    Characterizing the interaction conformation between T-cell receptors and epitopes with deep learning.Nature Machine Intelligence, 5(4):395–407, 2023

    Xingang Peng, Yipin Lei, Peiyuan Feng, Lemei Jia, Jianzhu Ma, Dan Zhao, and Jianyang Zeng. Characterizing the interaction conformation between T-cell receptors and epitopes with deep learning.Nature Machine Intelligence, 5(4):395–407, 2023. doi: 10.1038/ s42256-023-00634-4. URLhttps://doi.org/10.1038/s42256-023-00634-4

    work page doi:10.1038/s42256-023-00634-4 2023

  67. [67]

    T-Scan: a genome-wide method for the systematic discovery of T cell epitopes.Cell, 178(4):1016–1028,

    Tomasz Kula, Mohammad H Dezfulian, Charlotte I Wang, Nouran S Abdelfattah, Zachary C Hartman, Kai W Wucherpfennig, Herbert Kim Lyerly, and Stephen J Elledge. T-Scan: a genome-wide method for the systematic discovery of T cell epitopes.Cell, 178(4):1016–1028,

  68. [68]

    URLhttps://doi.org/10.1016/j.cell.2019

    doi: 10.1016/j.cell.2019.07.009. URLhttps://doi.org/10.1016/j.cell.2019. 07.009

    work page doi:10.1016/j.cell.2019.07.009 2019

  69. [69]

    Ragul Gowthaman and Brian G. Pierce. TCR3d: The T cell receptor structural repertoire database.Bioinformatics, 35(24):5323–5325, 2019. doi: 10.1093/bioinformatics/btz517. URL https://doi.org/10.1093/bioinformatics/btz517. 15

    work page doi:10.1093/bioinformatics/btz517 2019

  70. [70]

    TCR3d 2.0: expanding the T cell receptor structure database with new structures, tools and interactions.Nucleic Acids Research, 53(D1):D604–D608, 2025

    Valerie Lin, Melyssa Cheung, Ragul Gowthaman, Maya Eisenberg, Brian M Baker, and Brian G Pierce. TCR3d 2.0: expanding the T cell receptor structure database with new structures, tools and interactions.Nucleic Acids Research, 53(D1):D604–D608, 2025. doi: 10.1093/nar/gkae840. URLhttps://doi.org/10.1093/nar/gkae840

    work page doi:10.1093/nar/gkae840 2025

  71. [71]

    Jan W Gratama, Joost WJ van Esser, Cor HJ Lamers, Claire Tournay, Bob Lowenberg, Rein- der LH Bolhuis, and Jan J Cornelissen. Tetramer-based quantification of cytomegalovirus (CMV)–specific CD8+ T lymphocytes in T-cell–depleted stem cell grafts and after transplan- tation may identify patients at risk for progressive CMV infection.Blood, The Journal of th...

    work page doi:10.1182/blood.v98.5.1358 2001

  72. [72]

    Rinalmo: General-purpose rna language models can generalize well on structure prediction tasks.Na- ture Communications, 16(1):5671, 2025

    Rafael Josip Peni ´c, Tin Vla ˇsi´c, Roland G Huber, Yue Wan, and Mile ˇSiki´c. Rinalmo: General-purpose rna language models can generalize well on structure prediction tasks.Na- ture Communications, 16(1):5671, 2025. doi: 10.1038/s41467-025-60872-5. URLhttps: //doi.org/10.1038/s41467-025-60872-5

    work page doi:10.1038/s41467-025-60872-5 2025

  73. [73]

    DNABERT: pre-trained Bidirectional Encoder Representations from Transformers model for DNA-language in genome

    Yanrong Ji, Zhihan Zhou, Han Liu, and Ramana V Davuluri. DNABERT: pre-trained bidi- rectional encoder representations from transformers model for DNA-language in genome. Bioinformatics, 37(15):2112–2120, 2021. doi: 10.1093/bioinformatics/btab083. URLhttps: //doi.org/10.1093/bioinformatics/btab083

    work page doi:10.1093/bioinformatics/btab083 2021

  74. [74]

    Effective gene expression prediction from sequence by integrating long-range interac- tions.Nature methods, 18(10):1196–1203, 2021

    ˇZiga Avsec, Vikram Agarwal, Daniel Visentin, Joseph R Ledsam, Agnieszka Grabska- Barwinska, Kyle R Taylor, Yannis Assael, John Jumper, Pushmeet Kohli, and David R Kel- ley. Effective gene expression prediction from sequence by integrating long-range interac- tions.Nature methods, 18(10):1196–1203, 2021. doi: 10.1038/s41592-021-01252-x. URL https://doi.or...

    work page doi:10.1038/s41592-021-01252-x 2021

  75. [75]

    Epigept: a pretrained transformer-based language model for context-specific human epigenomics.Genome Biology, 25(1):1–30, 2024

    Zijing Gao, Qiao Liu, Wanwen Zeng, Rui Jiang, and Wing Hung Wong. Epigept: a pretrained transformer-based language model for context-specific human epigenomics.Genome Biology, 25(1):1–30, 2024. doi: 10.1186/s13059-024-03449-7. URLhttps://doi.org/10.1186/ s13059-024-03449-7

    work page doi:10.1186/s13059-024-03449-7 2024

  76. [76]

    ˇZiga Avsec, Natasha Latysheva, Jun Cheng, Guido Novati, Kyle R. Taylor, Tom Ward, Clare Bycroft, Lauren Nicolaisen, Eirini Arvaniti, Joshua Pan, Raina Thomas, Vincent Dutordoir, Matteo Perino, Soham De, Alexander Karollus, Adam Gayoso, Toby Sargeant, Anne Mot- tram, Lai Hong Wong, Pavol Drot´ar, Adam Kosiorek, Andrew Senior, Richard Tanburn, Tay- lor App...

    work page doi:10.1038/s41586-025-10014-0 2026

  77. [77]

    Mann, Michael Irvin, Defne G

    Maxim Zvyagin, Alexander Brace, Kyle Hippe, Yuntian Deng, Bin Zhang, Cindy Orozco Bo- horquez, Austin Clyde, Bharat Kale, Danilo Perez-Rivera, Heng Ma, Carla M. Mann, Michael Irvin, Defne G. Ozgulbas, Natalia Vassilieva, James Gregory Pauloski, Logan Ward, Valerie Hayot-Sasson, Murali Emani, Sam Foreman, Zhen Xie, Diangen Lin, Maulik Shukla, Weili Nie, Jo...

    work page doi:10.1177/10943420231201154 2023

  78. [78]

    Dinan, David C

    Maciej Wiatrak, Ramon Vinas Torne, Maria Ntemourtsidou, Adam M. Dinan, David C. Abel- son, Divya Arora, Maria Brbic, Aaron Weimann, and Rodrigo Andres Floto. A contextualised protein language model reveals the functional syntax of bacterial evolution.bioRxiv, 2025. doi: 10.1101/2025.07.20.665723. URLhttps://doi.org/10.1101/2025.07.20.665723

    work page doi:10.1101/2025.07.20.665723 2025

  79. [79]

    Li, Yepeng Huang, Marissa Sumathipala, Man Qing Liang, Alberto Valdeolivas, Ashwin N

    Michelle M. Li, Yepeng Huang, Marissa Sumathipala, Man Qing Liang, Alberto Valdeolivas, Ashwin N. Ananthakrishnan, Katherine Liao, Daniel Marbach, and Marinka Zitnik. Contextual AI models for single-cell protein biology.Nature Methods, 21(8):1546–1557, 2024. doi: 10.1038/s41592-024-02341-3. URLhttps://doi.org/10.1038/s41592-024-02341-3. 16

    work page doi:10.1038/s41592-024-02341-3 2024

  80. [80]

    Cornman, Elizabeth H

    Yunha Hwang, Andre L. Cornman, Elizabeth H. Kellogg, Sergey Ovchinnikov, and Pe- ter R. Girguis. Genomic language model predicts protein co-regulation and function.Na- ture Communications, 15(1):2880, 2024. doi: 10.1038/s41467-024-46947-9. URLhttps: //doi.org/10.1038/s41467-024-46947-9

    work page doi:10.1038/s41467-024-46947-9 2024

Showing first 80 references.