pith. sign in

arxiv: 2606.05474 · v1 · pith:MKZACZWFnew · submitted 2026-06-03 · 🧬 q-bio.BM · cs.LG

AlloGen: Conformation-Selective Binder Generation with Differential State Scoring

Hanqun Cao , Zachary Quinn , Aastha Pal , Sumi Kimura , Jingjie Zhang , Pheng Ann Heng , Pranam Chatterjee This is my paper

Pith reviewed 2026-06-28 03:16 UTC · model grok-4.3

classification 🧬 q-bio.BM cs.LG
keywords protein binder designconformational selectivityallosteric proteinsgraph transformerde novo peptide designstate discriminationbinder generation
0
0 comments X

The pith

AlloGen uses a learned scorer to generate protein binders selective for specific conformations.

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

Protein binder design has optimized for affinity but left conformational selectivity unaddressed, even though many targets like kinases require binders that engage only one state. AlloGen decouples backbone generation from a state-selectivity scorer Q_theta, an SE(3)-invariant interface graph transformer trained first on geometry then on conformational discrimination. The scorer is differentiable and generator-agnostic, so it can rerank or guide outputs from any existing tool without retraining. Benchmarks across protein families show it produces binders that prefer the desired state and reject alternatives. Experiments on calmodulin confirm the signals yield physical peptides that bind the holo form but show no detectable binding to the apo form.

Core claim

AlloGen decouples backbone generation from a learned state-selectivity scorer Q_θ, an SE(3)-invariant interface graph transformer trained via a two-phase curriculum that first learns interface geometry before imposing conformational discrimination. Because Q_θ is fully differentiable and generator-agnostic, it integrates with any backbone generator as a passive reranker or an active gradient-based guide without retraining. Across a diverse benchmark of proteins spanning multiple families and conformational mechanisms, AlloGen consistently identifies binders that preferentially recognize desired structural states while rejecting alternative conformations. Experimental validation on calmodulin

What carries the argument

The state-selectivity scorer Q_θ, an SE(3)-invariant interface graph transformer that scores binders for conformational preference after two-phase training on interface geometry and state discrimination.

If this is right

  • Any backbone generator can incorporate the scorer without retraining to enforce state selectivity.
  • Selective binders are produced across multiple protein families and conformational mechanisms.
  • Computational selectivity signals translate directly to physical binding specificity in laboratory tests.
  • Conformational selectivity becomes a learnable and designable property rather than an unaddressed requirement.

Where Pith is reading between the lines

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

  • Existing protein design pipelines could add the scorer module to gain selectivity for allosteric targets without redesigning the generator.
  • The approach suggests a route to functional specificity in binders for kinases, GPCRs, and nuclear receptors that would otherwise lack state discrimination.
  • The curriculum of geometry learning followed by discrimination training could be tested on other molecular properties such as binding site specificity or off-target avoidance.

Load-bearing premise

The learned state-selectivity scorer will generalize to new proteins and conformational mechanisms without retraining when used as a passive reranker or gradient guide with arbitrary backbone generators.

What would settle it

On a held-out set of proteins, the generated binders show no preference or reverse preference for the desired conformation over the alternative, or the calmodulin peptides exhibit detectable binding to the apo state in experiments.

Figures

Figures reproduced from arXiv: 2606.05474 by Aastha Pal, Hanqun Cao, Jingjie Zhang, Pheng Ann Heng, Pranam Chatterjee, Sumi Kimura, Zachary Quinn.

Figure 1
Figure 1. Figure 1: AlloGen pipeline. A frozen generator produces K binder backbones conditioned on the goal state X1 (holo, blue); the trained scorer Qθ evaluates each candidate against both X1 and the undesired state X0 (apo, red), and returns the top candidate Yˆ by selectivity margin ∆Q = Qθ(X1 , Y ) − Qθ(X0 , Y ). Qθ is trained independently and plugs into any backbone generator without retraining. over three training se… view at source ↗
Figure 2
Figure 2. Figure 2: Qθ selectivity performance. (a)Qθ scoring performance ablation by data augmentation strategies and two phases; (b) Qθ scoring performance ablation per target by different features. positive, indicating that conformational selectivity is recoverable only from a representation trained on paired apo and holo geometry [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qθ conformational selectivity and CaM selectivity design. (a) Cross target selectivity matrix between Source Target (binders generated for) and the Reference Target (scored against). Zero values (0.00) have been omitted.; (b) Holo vs. apo Qθ scores for 50 vanilla designs per target; (c) Selectivity-based design on CaM. Two case binders (orange) shown against the apo (1st and 3rd panels) and holo (2nd and 4… view at source ↗
Figure 4
Figure 4. Figure 4: Generation benchmark on CaM. (a) Consensus selectivity S¯ cons across 15 generation × guidance approaches. (b) Selectivity vs. design success rate (designable × selective) across all generator × guidance combinations. We then applied the full pipeline to CaM, a stringent test because its ∼30 Å apo-to-holo rear￾rangement on Ca2+ binding opens a hydrophobic peptide-binding cleft that is occluded in the apo s… view at source ↗
Figure 5
Figure 5. Figure 5: Experimental validation workflow for conformationally selective peptide binding to [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Protein binder design has largely optimized for affinity alone, leaving conformational selectivity unaddressed: for allosteric targets such as kinases, nuclear receptors, and GPCRs, a binder that engages both active and inactive states provides no functional specificity regardless of how tightly it binds. We introduce AlloGen, a modular framework that decouples backbone generation from a learned state-selectivity scorer $Q_\theta$, an SE(3)-invariant interface graph transformer trained via a two-phase curriculum that first learns interface geometry before imposing conformational discrimination. Because $Q_\theta$ is fully differentiable and generator-agnostic, it integrates with any backbone generator as a passive reranker or an active gradient-based guide without retraining. Across a diverse benchmark of proteins spanning multiple families and conformational mechanisms, AlloGen consistently identifies binders that preferentially recognize desired structural states while rejecting alternative conformations. Experimental validation on calmodulin further demonstrates that these computational selectivity signals translate to physical molecules, yielding de novo peptides that bind the desired holo conformation while exhibiting no detectable binding to the apo state. Together, these results establish conformational selectivity as a learnable property and provide a general framework for state-selective protein binder design.

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 / 2 minor

Summary. The manuscript introduces AlloGen, a modular framework for conformation-selective protein binder design. It decouples backbone generation from a learned state-selectivity scorer Q_θ (an SE(3)-invariant interface graph transformer) trained via a two-phase curriculum on interface geometry followed by conformational discrimination. The scorer is presented as fully differentiable and generator-agnostic, allowing integration with arbitrary backbone generators as a passive reranker or active gradient guide without retraining. The central claims are consistent identification of state-preferential binders across a diverse benchmark spanning multiple protein families and conformational mechanisms, plus experimental validation on calmodulin yielding de novo peptides that bind the holo but not the apo state.

Significance. If the generalization of Q_θ holds, the work would establish conformational selectivity as a learnable, modular property and provide a practical framework for state-selective design targeting allosteric proteins such as kinases, nuclear receptors, and GPCRs. The experimental translation on calmodulin would be a notable strength, as would the generator-agnostic architecture if supported by cross-generator and cross-mechanism evidence.

major comments (2)
  1. [Abstract] Abstract: the claim that AlloGen 'consistently identifies binders that preferentially recognize desired structural states' across a 'diverse benchmark of proteins spanning multiple families and conformational mechanisms' is load-bearing for the generalization assertion, yet the abstract supplies no quantitative metrics, success rates, baseline comparisons, error analysis, or details on training-set exclusion criteria and test-protein mechanisms.
  2. [Abstract] Abstract: the assertion that the two-phase curriculum produces a scorer that 'generalizes to new proteins and conformational mechanisms without retraining' when used with arbitrary backbone generators rests on the untested assumption that the discrimination phase isolates conformational state rather than protein-specific interface statistics; no ablation, cross-validation on held-out mechanisms (e.g., kinases or GPCRs), or transfer results are referenced to support this.
minor comments (2)
  1. [Abstract] The abstract states 'no detectable binding to the apo state' for the calmodulin peptides but provides no experimental details, controls, or quantitative binding data.
  2. [Abstract] Notation Q_θ is introduced without a brief definition of its input representation or output scale in the abstract.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract. The comments highlight opportunities to strengthen the presentation of quantitative results and generalization evidence. We will revise the abstract accordingly while preserving its brevity. Point-by-point responses follow.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that AlloGen 'consistently identifies binders that preferentially recognize desired structural states' across a 'diverse benchmark of proteins spanning multiple families and conformational mechanisms' is load-bearing for the generalization assertion, yet the abstract supplies no quantitative metrics, success rates, baseline comparisons, error analysis, or details on training-set exclusion criteria and test-protein mechanisms.

    Authors: We agree the abstract should be more informative. In revision we will incorporate concise quantitative metrics (e.g., success rates and baseline comparisons), a brief statement on training-set exclusion, and high-level details on the test-protein mechanisms and families. Full error analysis, per-protein breakdowns, and exclusion criteria remain in the Results and Methods sections. revision: yes

  2. Referee: [Abstract] Abstract: the assertion that the two-phase curriculum produces a scorer that 'generalizes to new proteins and conformational mechanisms without retraining' when used with arbitrary backbone generators rests on the untested assumption that the discrimination phase isolates conformational state rather than protein-specific interface statistics; no ablation, cross-validation on held-out mechanisms (e.g., kinases or GPCRs), or transfer results are referenced to support this.

    Authors: The curriculum first trains on interface geometry across diverse interfaces and then adds conformational discrimination; the benchmark evaluates the resulting scorer on held-out proteins spanning multiple families and mechanisms without retraining. We will revise the abstract to explicitly reference the relevant benchmark results and figures that demonstrate cross-protein transfer. Explicit ablations that further isolate the discrimination phase from protein-specific statistics are not currently reported; if the editor requests, we can add a short supplementary analysis in revision. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external training data and modular application.

full rationale

The paper's core construction trains Q_θ (an SE(3)-invariant graph transformer) via two-phase curriculum on interface geometry then conformational discrimination using external structural data, then applies the fixed scorer as a generator-agnostic reranker or gradient guide. No step equates a claimed prediction or selectivity signal to its own fitted inputs by definition, nor does any load-bearing premise reduce to a self-citation chain. The benchmark results and calmodulin validation are presented as downstream empirical outcomes rather than tautological outputs. Generalization to unseen proteins is an empirical claim (with associated risk) but does not create a definitional loop within the reported derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; full text would be required to enumerate training hyperparameters, data assumptions, or any new postulated components.

pith-pipeline@v0.9.1-grok · 5756 in / 1149 out tokens · 38611 ms · 2026-06-28T03:16:49.642044+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

49 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    Novel aspects of calmodulin target recognition and activation.European Journal of Biochemistry, 270(3):404–414, 2003

    Stefan W Vetter and Estelle Leclerc. Novel aspects of calmodulin target recognition and activation.European Journal of Biochemistry, 270(3):404–414, 2003

    2003

  2. [2]

    Protein conformational switches: from nature to design

    Jeung-Hoi Ha and Stewart N Loh. Protein conformational switches: from nature to design. Chemistry–A European Journal, 18(26):7984–7999, 2012

    2012

  3. [3]

    Conformational selection in protein binding and function

    Thomas R Weikl and Fabian Paul. Conformational selection in protein binding and function. Protein Science, 23(11):1508–1518, 2014

    2014

  4. [4]

    Introduction to protein ensembles and allostery, 2016

    Ruth Nussinov. Introduction to protein ensembles and allostery, 2016

    2016

  5. [5]

    Allostery and population shift in drug discovery.Current opinion in pharmacology, 10(6):715–722, 2010

    Gozde Kar, Ozlem Keskin, Attila Gursoy, and Ruth Nussinov. Allostery and population shift in drug discovery.Current opinion in pharmacology, 10(6):715–722, 2010

    2010

  6. [6]

    Conformational analysis of the dfg-out kinase motif and biochemical profiling of structurally validated type ii inhibitors.Journal of medicinal chemistry, 58(1):466–479, 2015

    RSK Vijayan, Peng He, Vivek Modi, Krisna C Duong-Ly, Haiching Ma, Jeffrey R Peterson, Roland L Dunbrack Jr, and Ronald M Levy. Conformational analysis of the dfg-out kinase motif and biochemical profiling of structurally validated type ii inhibitors.Journal of medicinal chemistry, 58(1):466–479, 2015

    2015

  7. [7]

    Small molecule modulation of nuclear receptor conformational dynamics: implications for function and drug discovery.Molecular pharma- cology, 83(1):1–8, 2013

    Douglas J Kojetin and Thomas P Burris. Small molecule modulation of nuclear receptor conformational dynamics: implications for function and drug discovery.Molecular pharma- cology, 83(1):1–8, 2013

    2013

  8. [8]

    Functional dynamics of g protein-coupled receptors reveal new routes for drug discovery

    Paolo Conflitti, Edward Lyman, Mark SP Sansom, Peter W Hildebrand, Hugo Gutiérrez-de Terán, Paolo Carloni, T Bertie Ansell, Shuguang Yuan, Patrick Barth, Anne S Robinson, et al. Functional dynamics of g protein-coupled receptors reveal new routes for drug discovery. Nature Reviews Drug Discovery, 24(4):251–275, 2025

    2025

  9. [9]

    De novo design of bioactive protein switches.Nature, 572(7768):205–210, 2019

    Robert A Langan, Scott E Boyken, Andrew H Ng, Jennifer A Samson, Galen Dods, Alexandra M Westbrook, Taylor H Nguyen, Marc J Lajoie, Zibo Chen, Stephanie Berger, et al. De novo design of bioactive protein switches.Nature, 572(7768):205–210, 2019

    2019

  10. [10]

    Target sequence-conditioned design of peptide binders using masked language modeling.Nature Biotechnology, pages 1–9, 2025

    Leo Tianlai Chen, Zachary Quinn, Madeleine Dumas, Christina Peng, Lauren Hong, Moi- ses Lopez-Gonzalez, Alexander Mestre, Rio Watson, Sophia Vincoff, Lin Zhao, et al. Target sequence-conditioned design of peptide binders using masked language modeling.Nature Biotechnology, pages 1–9, 2025

    2025

  11. [11]

    De novo design of peptide binders to con- formationally diverse targets with contrastive language modeling.Science Advances, 11(4): eadr8638, 2025

    Suhaas Bhat, Kalyan Palepu, Lauren Hong, Joey Mao, Tianzheng Ye, Rema Iyer, Lin Zhao, Tianlai Chen, Sophia Vincoff, Rio Watson, et al. De novo design of peptide binders to con- formationally diverse targets with contrastive language modeling.Science Advances, 11(4): eadr8638, 2025

    2025

  12. [12]

    Peptune: De novo generation of thera- peutic peptides with multi-objective-guided discrete diffusion

    Sophia Tang, Yinuo Zhang, and Pranam Chatterjee. Peptune: De novo generation of thera- peutic peptides with multi-objective-guided discrete diffusion. InForty-second International Conference on Machine Learning, 2025. URL https://openreview.net/forum?id= FQoy1Y1Hd8

    2025

  13. [13]

    Soapia: Siamese-guided generation of off target-avoiding protein interactions with high target affinity

    Sophia Vincoff, Oscar Davis, Ismail Ilkan Ceylan, Alexander Tong, Joey Bose, and Pranam 16 Chatterjee. Soapia: Siamese-guided generation of off target-avoiding protein interactions with high target affinity. InICML 2025 Generative AI and Biology (GenBio) Workshop, 2025

    2025

  14. [14]

    GLID$^2$e: A gradient-free lightweight fine-tune approach for discrete biological sequence design

    Hanqun Cao, Haosen Shi, Chenyu Wang, Sinno Jialin Pan, and Pheng-Ann Heng. GLID$^2$e: A gradient-free lightweight fine-tune approach for discrete biological sequence design. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025. URL https://openreview.net/forum?id=AHjspi4R22

    2025

  15. [15]

    TD3b: Transition-directed discrete diffusion for allosteric binder gen- eration

    Hanqun Cao, Aastha Pal, Sophia Tang, Yinuo Zhang, Jingjie Zhang, Pheng-Ann Heng, and Pranam Chatterjee. TD3b: Transition-directed discrete diffusion for allosteric binder gen- eration. InForty-Third International Conference on Machine Learning, 2026. URL https: //openreview.net/forum?id=r2nuhKd6vr

    2026

  16. [16]

    De novo design of protein structure and function with rfdiffusion.Nature, 620(7976):1089–1100, 2023

    Joseph L Watson, David Juergens, Nathaniel R Bennett, Brian L Trippe, Jason Yim, Helen E Eisenach, Woody Ahern, Andrew J Borst, Robert J Ragotte, Lukas F Milles, et al. De novo design of protein structure and function with rfdiffusion.Nature, 620(7976):1089–1100, 2023

    2023

  17. [17]

    Pxdesign: Fast, modular, and accurate de novo design of protein binders.bioRxiv, 2025

    Protenix Team, Milong Ren, Jinyuan Sun, Jiaqi Guan, Cong Liu, Chengyue Gong, Yuzhe Wang, Lan Wang, Qixu Cai, Xinshi Chen, et al. Pxdesign: Fast, modular, and accurate de novo design of protein binders.bioRxiv, 2025

    2025

  18. [18]

    Bronstein, Martin Steinegger, Emine Kucukbenli, Arash Vahdat, and Karsten Kreis

    Kieran Didi, Zuobai Zhang, Guoqing Zhou, Danny Reidenbach, Zhonglin Cao, Sooyoung Cha, Tomas Geffner, Christian Dallago, Jian Tang, Michael M. Bronstein, Martin Steinegger, Emine Kucukbenli, Arash Vahdat, and Karsten Kreis. Scaling atomistic protein binder design with generative pretraining and test-time compute. InThe Fourteenth International Confer- enc...

    2026

  19. [19]

    Boltzgen: Toward universal binder design.bioRxiv, 2025

    Hannes Stark, Felix Faltings, MinGyu Choi, Yuxin Xie, Eunsu Hur, Timothy O’Donnell, Anton Bushuiev, Talip Uçar, Saro Passaro, Weian Mao, et al. Boltzgen: Toward universal binder design.bioRxiv, 2025

    2025

  20. [20]

    One-shot design of functional protein binders with bindcraft.Nature, 646(8084):483–492, 2025

    Martin Pacesa, Lennart Nickel, Christian Schellhaas, Joseph Schmidt, Ekaterina Pyatova, Lucas Kissling, Patrick Barendse, Jagrity Choudhury, Srajan Kapoor, Ana Alcaraz-Serna, et al. One-shot design of functional protein binders with bindcraft.Nature, 646(8084):483–492, 2025

    2025

  21. [21]

    Highly accurate protein structure prediction with alphafold.nature, 596(7873):583–589, 2021

    John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ron- neberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Žídek, Anna Potapenko, et al. Highly accurate protein structure prediction with alphafold.nature, 596(7873):583–589, 2021

    2021

  22. [22]

    Evolutionary-scale prediction of atomic-level protein structure with a language model.Science, 379(6637):1123–1130, 2023

    Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Nikita Smetanin, Robert Verkuil, Ori Kabeli, Yaniv Shmueli, et al. Evolutionary-scale prediction of atomic-level protein structure with a language model.Science, 379(6637):1123–1130, 2023

    2023

  23. [23]

    Prodigy: a web server for predicting the binding affinity of protein–protein complexes

    Li C Xue, João Pglm Rodrigues, Panagiotis L Kastritis, Alexandre Mjj Bonvin, and Anna Van- gone. Prodigy: a web server for predicting the binding affinity of protein–protein complexes. Bioinformatics, 32(23):3676–3678, 2016. 17

    2016

  24. [24]

    Molecular and structural basis of target recognition by calmodulin.Annual review of biophysics and biomolecular structure, 24:85–116, 1995

    Anna Crivici and Mitsuhiko Ikura. Molecular and structural basis of target recognition by calmodulin.Annual review of biophysics and biomolecular structure, 24:85–116, 1995

    1995

  25. [25]

    Dockq: a quality measure for protein-protein docking models

    Sankar Basu and Björn Wallner. Dockq: a quality measure for protein-protein docking models. PloS one, 11(8):e0161879, 2016

    2016

  26. [26]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017

    2017

  27. [27]

    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

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  28. [28]

    Robust deep learning–based protein sequence design using proteinmpnn.Science, 378(6615):49–56, 2022

    Justas Dauparas, Ivan Anishchenko, Nathaniel Bennett, Hua Bai, Robert J Ragotte, Lukas F Milles, Basile IM Wicky, Alexis Courbet, Rob J de Haas, Neville Bethel, et al. Robust deep learning–based protein sequence design using proteinmpnn.Science, 378(6615):49–56, 2022

    2022

  29. [29]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. In International Conference on Learning Representations, 2021. URL https://openreview. net/forum?id=PxTIG12RRHS

    2021

  30. [30]

    Bayesian learning via stochastic gradient langevin dynamics

    Max Welling and Yee W Teh. Bayesian learning via stochastic gradient langevin dynamics. InProceedings of the 28th international conference on machine learning (ICML-11), pages 681–688, 2011

    2011

  31. [31]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021

    2021

  32. [32]

    Practical and asymptotically exact conditional sampling in diffusion models.Advances in Neural Information Processing Systems, 36:31372–31403, 2023

    Luhuan Wu, Brian Trippe, Christian Naesseth, David Blei, and John P Cunningham. Practical and asymptotically exact conditional sampling in diffusion models.Advances in Neural Information Processing Systems, 36:31372–31403, 2023

    2023

  33. [33]

    PPDiff: Diffusing in hybrid sequence-structure space for protein-protein complex design

    Zhenqiao Song, Tianxiao Li, Lei Li, and Martin Renqiang Min. PPDiff: Diffusing in hybrid sequence-structure space for protein-protein complex design. InForty-second International Conference on Machine Learning, 2025. URL https://openreview.net/forum?id= gTYzm0Qchr

    2025

  34. [34]

    moppit-v3: Motif-specific peptides generated via multi-objective-guided discrete flow matching

    Tong Chen, Zachary Quinn, Yinuo Zhang, and Pranam Chatterjee. moppit-v3: Motif-specific peptides generated via multi-objective-guided discrete flow matching. InNeurIPS 2025 AI for Science Workshop

    2025

  35. [35]

    arXiv preprint arXiv:2509.25171 , year=

    Sophia Tang, Yuchen Zhu, Molei Tao, and Pranam Chatterjee. Tr2-d2: Tree search guided trajectory-aware fine-tuning for discrete diffusion.arXiv preprint arXiv:2509.25171, 2025

    work page arXiv 2025

  36. [36]

    Areuredi: Annealed rectified updates for refining discrete flows with multi-objective guidance.arXiv preprint arXiv:2510.00352, 2025

    Tong Chen, Yinuo Zhang, and Pranam Chatterjee. Areuredi: Annealed rectified updates for refining discrete flows with multi-objective guidance.arXiv preprint arXiv:2510.00352, 2025

    work page arXiv 2025

  37. [37]

    Gumbel-softmax score and flow matching for discrete biological sequence generation

    Sophia Tang, Yinuo Zhang, Alexander Tong, and Pranam Chatterjee. Gumbel-softmax score and flow matching for discrete biological sequence generation. 2025. 18

    2025

  38. [38]

    The foldx web server: an online force field.Nucleic acids research, 33(suppl_2): W382–W388, 2005

    Joost Schymkowitz, Jesper Borg, Francois Stricher, Robby Nys, Frederic Rousseau, and Luis Serrano. The foldx web server: an online force field.Nucleic acids research, 33(suppl_2): W382–W388, 2005

    2005

  39. [39]

    Learning inverse folding from millions of predicted structures

    Chloe Hsu, Robert Verkuil, Jason Liu, Zeming Lin, Brian Hie, Tom Sercu, Adam Lerer, and Alexander Rives. Learning inverse folding from millions of predicted structures. InInterna- tional conference on machine learning, pages 8946–8970. PMLR, 2022

    2022

  40. [40]

    Protein docking model evaluation by graph neural networks.Frontiers in Molecular Biosciences, 8:647915, 2021

    Xiao Wang, Sean T Flannery, and Daisuke Kihara. Protein docking model evaluation by graph neural networks.Frontiers in Molecular Biosciences, 8:647915, 2021

    2021

  41. [41]

    Deeprank-gnn-esm: a graph neural network for scoring protein–protein models using protein language model.Bioinformatics advances, 4 (1):vbad191, 2024

    Xiaotong Xu and Alexandre MJJ Bonvin. Deeprank-gnn-esm: a graph neural network for scoring protein–protein models using protein language model.Bioinformatics advances, 4 (1):vbad191, 2024

    2024

  42. [42]

    Alphafold meets flow matching for gener- ating protein ensembles

    Bowen Jing, Bonnie Berger, and Tommi Jaakkola. Alphafold meets flow matching for gener- ating protein ensembles. InForty-first International Conference on Machine Learning, 2024. URLhttps://openreview.net/forum?id=rs8Sh2UASt

    2024

  43. [43]

    Afsample2 predicts multiple conformations and ensem- bles with alphafold2.Communications biology, 8(1):373, 2025

    Yogesh Kalakoti and Björn Wallner. Afsample2 predicts multiple conformations and ensem- bles with alphafold2.Communications biology, 8(1):373, 2025

    2025

  44. [44]

    Accurate structure prediction of biomolecular interactions with alphafold 3.Nature, 630(8016):493–500, 2024

    Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J Ballard, Joshua Bambrick, et al. Accurate structure prediction of biomolecular interactions with alphafold 3.Nature, 630(8016):493–500, 2024

    2024

  45. [45]

    A comparison of successful and failed protein interface designs highlights the challenges of designing buried hydrogen bonds.Protein Science, 22(1):74–82, 2013

    P Benjamin Stranges and Brian Kuhlman. A comparison of successful and failed protein interface designs highlights the challenges of designing buried hydrogen bonds.Protein Science, 22(1):74–82, 2013

    2013

  46. [46]

    Computational design of proteins targeting the conserved stem region of influenza hemagglutinin.Science, 332(6031):816–821, 2011

    Sarel J Fleishman, Timothy A Whitehead, Damian C Ekiert, Cyrille Dreyfus, Jacob E Corn, Eva-Maria Strauch, Ian A Wilson, and David Baker. Computational design of proteins targeting the conserved stem region of influenza hemagglutinin.Science, 332(6031):816–821, 2011

    2011

  47. [47]

    Automated design of specificity in molecular recog- nition.Nature structural biology, 10(1):45–52, 2003

    James J Havranek and Pehr B Harbury. Automated design of specificity in molecular recog- nition.Nature structural biology, 10(1):45–52, 2003

    2003

  48. [48]

    State-specific peptide design targeting g protein- coupled receptors.Journal of Chemical Information and Modeling, 65(20):11425–11438, 2025

    Yang Xue, Hong Wang, Jun Li, Jianguo Hu, Zhiyuan Chen, Zhi Zheng, Lihang Liu, Kunrui Zhu, Jingzhou He, Huanzhang Gong, et al. State-specific peptide design targeting g protein- coupled receptors.Journal of Chemical Information and Modeling, 65(20):11425–11438, 2025

    2025

  49. [49]

    Generating functional and multistate proteins with a multimodal diffusion transformer.bioRxiv, 2025

    Bowen Jing, Anna Sappington, Mihir Bafna, Ravi Shah, Adrina Tang, Rohith Krishna, Adam Klivans, Daniel J Diaz, and Bonnie Berger. Generating functional and multistate proteins with a multimodal diffusion transformer.bioRxiv, 2025. 19 Supplementary Information S1 Related Work S1.1 Structure-Based Binder Design Deep generative models have transformed struct...

    work page arXiv 2025