Pith. sign in

REVIEW 4 major objections 5 minor 52 references

DS-ProGen: A Dual-Structure Deep Language Model for Functional Protein Design

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A dual-structure language model for inverse protein folding recovers 61.47% of native amino acids by conditioning on both backbone geometry and molecular surface features.

desk verdict A genuinely new dual-structure architecture for inverse protein folding, but the 61.47% SOTA claim is not supported by the experiments as written due to unfair baselines and an undefined test subset. read the letter →

arxiv 2505.12511 v1 pith:JOEB5SB2 submitted 2025-05-18 cs.CL

Yanting Li , Jiyue Jiang , Zikang Wang , Ziqian Lin , Dongchen He , Yuheng Shan , Yanruisheng Shao , Jiayi Li
show 6 more authors
Xiangyu Shi Jiuming Wang Yanyu Chen Yimin Fan Han Li Yu Li
This is my paper · ORCID
classification cs.CL
keywords inverseproteinfoldingdesigndual-structureencodingbackbonegeometrymolecularsurfacefeaturesautoregressivelanguagemodelsequencerecoveryPRIDEbenchmark
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that inverse protein folding — choosing an amino acid sequence that folds into a given three-dimensional structure — is better solved when the structure is read twice: once through backbone coordinates and once through the molecular surface. To test this, it builds DS-ProGen, an autoregressive language model that sums a backbone-geometry embedding and a surface-chemistry embedding into a single structural condition and feeds that condition to every next-amino-acid prediction. On the PRIDE benchmark the model recovers 61.47% of native residues, and its ablations show both channels matter: dropping the surface encoder lowers recovery to 50.71%, while dropping the backbone encoder lowers it to 31.06%. The paper concludes that surface features and backbone features are complementary constraints, not redundant views, for sequence design.

What carries the argument

The central object is the dual-structure fusion embedding $R = B + S$, a simple sum of two structural embeddings that becomes the conditioning context for every amino-acid prediction. $B$ comes from a backbone encoder: N, C, and Cα coordinates, dihedral angles, pairwise distances, and local orientation vectors updated by four geometric-vector-perceptron layers and a transformer, producing one vector per residue. $S$ comes from a surface encoder: a point cloud of the solvent-accessible surface built with a smooth distance function and van der Waals radii, each point carrying atom-type one-hots, multiscale mean and Gaussian curvatures, and the chemical identity and distance of its 16 nearest atoms, patched, message-passed, and pooled to one vector per residue. The mechanism does its work through the additive fusion: the decoder never sees the modalities separately, so every next-token choice must satisfy global topology and local surface chemistry simultaneously.

What would settle it

Re-run the benchmark with every baseline fine-tuned on the same PRIDE training data, decoded with the same temperature 0.1 and top-k 10, and include the missing published baseline; if a backbone-only baseline trained on the same surface-filtered subset reaches 61.47%, or the margin over the next-best method shrinks to noise, the dual-structure fusion is not the source of the reported gain.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that fusing backbone and surface information into one structural prior improves inverse-folding accuracy beyond either modality alone. The model extracts a backbone embedding from N, C, and Cα coordinates using rotation-equivariant geometric vector perceptrons and a transformer, extracts a surface embedding from surface point clouds with atom types, multiscale curvatures, and local chemical neighborhoods, then forms the structural condition $R = B + S$ and prepends it to the sequence tokens of an autoregressive decoder. Trained on the PRIDE benchmark with pretraining on large structure-sequence collections, this dual-conditioned decoder reaches 61.47% average recovery and 63.50% on short chains, while generating sequences whose predicted structures stay close to the target fold. The paper presents this as evidence that internal fold and external chemistry jointly constrain the native sequence, and that a language-model decoder can exploit both.

Load-bearing premise

The load-bearing premise is that the headline 61.47% recovery rate comes from a fair, apples-to-apples comparison: the baselines are evaluated under the same fine-tuning, decoding, and test-set conditions as DS-ProGen, even though the surface model was trained on a chemically filtered subset and at least one published PRIDE baseline is missing from the main table.

Editorial extensions

If this is right

  • If the 61.47% figure holds in a fair comparison, inverse folding no longer requires choosing between backbone-only and surface-only designs; the two channels are complementary and can be combined additively.
  • The surface channel's largest contribution appears on short chains (63.50% versus 43.75% for the best single-modality baseline), suggesting surface chemistry is most informative where the backbone graph is too sparse for geometry-only encoders.
  • The ablations imply the two encoders carry different information: removing surface features costs roughly 11 percentage points of recovery, while removing backbone features costs roughly 30, so global fold is the primary constraint and surface chemistry is a local refinement.
  • The generated sequences are about 38% different from native sequences yet fold back to the target with low RMSD, indicating the model has learned a structure-to-sequence mapping rather than memorized training sequences.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A decisive test not in the paper: fine-tune a strong backbone-only model on exactly the same surface-filtered training subset and decode with the same temperature and top-k; if its recovery approaches 61.47%, the gain currently attributed to fusion is really due to training-subset differences.
  • The same dual-stream conditioning should transfer to neighboring tasks the paper only gestures at, such as RNA inverse folding or ligand-pocket redesign, where surface chemistry dominates the design constraints; a successful transfer would support the general mechanism rather than the specific benchmark.
  • Because the surface encoder is restricted to proteins containing only C, N, O, S, Se, and H atoms, the reported gain may partly reflect selection of chemically simple targets; extending the atom vocabulary to metals and modified residues would test whether the fusion idea survives on harder proteomes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes DS-ProGen, an autoregressive protein design model that combines a GVP-based backbone encoder (initialized from ESM-IF) with a surface point-cloud encoder and a ProGen2-small-initialized decoder. The central empirical claim is a 61.47% amino-acid recovery rate on the PRIDE benchmark, presented as state-of-the-art and as evidence that fusing backbone geometry with molecular surface features improves inverse protein folding. The paper also reports AlphaFold3-based structural fidelity metrics and qualitative ligand/ion design case studies.

Significance. If the empirical claims held under a fully controlled comparison, this would be a useful contribution: it would demonstrate that surface descriptors and backbone geometry can be fused in a language-model inverse-folding framework, with improved sequence recovery and plausible structural predictions. The architecture description is fairly detailed, and the data-leakage analysis in Appendix A.3 is a positive feature. However, the current evaluation does not support the state-of-the-art claim because the baseline comparison is not apples-to-apples: ESM-IF is not fine-tuned while other baselines are retrained, a published PRIDE baseline (iProFold) is absent from the main recovery table, and the test-set definition for the surface model is ambiguous.

major comments (4)
  1. [§4.3, Table 1] The comparison against ESM-IF is not controlled. ProteinMPNN and PiFold are retrained on the PRIDE training set, while ESM-IF is evaluated from released pretrained weights without fine-tuning; DS-ProGen, in contrast, is initialized from ProGen2-small and ESM-IF weights and fine-tuned on the PRIDE training set. Because ESM-IF's backbone encoder and language-model prior are directly used as the starting point, the 9.62-point gap (51.85 vs. 61.47) cannot be attributed to the dual-structure architecture unless a fine-tuned ESM-IF baseline is provided under identical training data, initialization, and decoding. Please add this control.
  2. [§4.1, Table 1] The test-set definition for the surface model is ambiguous. Section 4.1 states that the surface encoder is trained only on samples with available surface information, and Section 3.2 states that proteins containing atom types outside C, N, O, S, Se, and H are discarded. The paper never reports how many of the 504 PRIDE test proteins have valid surface data, nor whether Table 1 is computed on the full test set or on a surface-valid subset, nor what happens at inference when surface data are unavailable. If the reported 61.47% is on an easier subset, it is not comparable with baselines evaluated on the full test set. Please report the exact test subset and re-evaluate all baselines on that subset.
  3. [Appendix Fig. 7, Table 1] The main recovery table omits iProFold and iProFold-B even though both appear in the RMSD comparison of Appendix Figure 7. iProFold is a published PRIDE baseline, so the abstract and conclusion's 'current state-of-the-art' label cannot be verified without its recovery rate in Table 1 (or a documented reason for exclusion). Please add these numbers or explicitly state why they are not comparable.
  4. [§4.1, Table 4] There is an inconsistency in the pretraining-dataset size: Section 4.1 says the large pretraining set contains 40 million structure-sequence pairs and the surface pretraining set contains 80,000, while Table 4's caption says 'Pretraining on 4M structure-sequence pairs.' Since the ablation claims a roughly 19% gain from pretraining, the correct dataset sizes must be stated and the caption reconciled.
minor comments (5)
  1. [§4.3] The decoding protocol (temperature, top-k) is specified for DS-ProGen but not for the baselines; please report whether the same sampling settings were used for all models.
  2. [Table 2] The text notes that DS-ProGen has the lowest RMSD but a lower mean TM-Score than PiFold and ESM-IF; the discussion is appropriately cautious, but the comparison would be strengthened by reporting confidence intervals or per-sample paired statistics.
  3. [Figures 4, 10–12] The ligand/ion interaction evaluation is qualitative; a quantitative metric (e.g., binding-pocket RMSD or interaction fingerprint similarity) would be needed to support the claim of 'functional retention.'
  4. [Appendix Fig. 9] The label 'DS-ProGen-B' is used in the figure while the main text uses 'backbone-only'; please define this abbreviation.
  5. [General] The manuscript does not state whether code and trained models will be released, which is important for reproducibility of the reported architecture and benchmarks.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the SOTA claim is tested on an external benchmark; only minor non-load-bearing self-citations appear in related work.

full rationale

The central derivation is self-contained. DS-ProGen is trained on the external PRIDE benchmark (32,389 CATH proteins) and evaluated on the held-out 504 CAMEO test set; the recovery rate 61.47% is computed from Eq. (15) against held-out ground-truth sequences, not from any fitted parameter. The backbone encoder is initialized from ESM-IF and the decoder from ProGen2-small, but these are external pretrained weights used as transfer-learning initializations, and the new surface branch plus fine-tuning adds trainable content; no prediction reduces by construction to those inputs. The ablations (w/o s_enc, w/o b_enc, w/o pretraining) are run within the same framework and do not rename a fitted quantity. AlphaFold3 is an external structure-prediction tool; its use for evaluating structural fidelity does not input the paper's own predictions as ground truth. The only self-citations are UniGenX and the authors' bioinformatics surveys in Related Work, and none is load-bearing for the main result. Concerns raised by the comparison—ESM-IF not fine-tuned, iProFold absent from Table 1, and the undefined surface-valid test subset—are evaluation-fairness or correctness issues, not circularity under the specified patterns.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. It relies on standard deep learning and structural biology assumptions: that the smooth-distance-function surface with van der Waals radii is useful, that 16-neighbor chemical features and multiscale curvatures capture local context, that AlphaFold3 is an adequate evaluation surrogate, and that pretrained ProGen2 and ESM-IF weights transfer positively. Several hyperparameters, including surface point count, sampling temperature, and the six-atom filter, are hand-chosen and directly affect the reported recovery rate.

free parameters (6)
  • Number of surface points per protein = 8192 (sampled or padded)
    Section 3.2 caps the surface point cloud at 8,192 points, chosen for batching. This limit can discard surface detail and is a hand-selected value that affects the surface encoder output.
  • Sampling temperature = 0.1
    Section 4.1 sets inference temperature to 0.1. Lower temperatures typically increase recovery rate, so this choice can inflate the reported metric relative to baselines using other decoding strategies.
  • Top-k = 10
    Section 4.1 uses top-k sampling with k=10, which controls the diversity of generated sequences and directly affects recovery rate.
  • Curvature radii = 1, 2, 3, 5, 10 Å
    Section 3.2 computes multiscale curvatures across five hand-selected radii. Different radii would change the geometric descriptors and the surface encoder's representation.
  • K for KNN surface patches = not specified
    Section 3.2 says each patch contains K surface points, but the value of K is not reported, making the exact architecture irreproducible and the choice unverifiable.
  • Number of FPS center points g = not specified
    Section 3.2 uses farthest point sampling to select g centers for patches, but g is not reported. This is a free architectural choice.
assumptions (5)
  • domain assumption The molecular surface generated by Blinn's smooth distance function using van der Waals radii is a sufficient representation for inverse folding.
    Section 3.2 constructs surface point clouds with this method. If this surface model omits chemically relevant detail, the surface encoder's contribution is misestimated.
  • domain assumption AlphaFold3-predicted structures are accurate enough to evaluate whether designed sequences fold into the target.
    Section 4.2 uses AlphaFold3 to compute TM-Score and RMSD. The authors acknowledge in Limitations that prediction errors are not accounted for, so structural fidelity claims depend on this assumption.
  • domain assumption Pretrained weights from ProGen2-small and ESM-IF transfer to the dual-structure architecture and improve fine-tuning performance.
    Section 4.1 initializes the decoder and backbone encoder from these models. The ablation 'w/o pretraining' shows a large drop, so part of the reported gain is attributed to pretraining rather than the dual-structure design alone.
  • domain assumption The 80,000-protein surface pretraining set is diverse enough to support the backbone and surface encoders.
    Section 4.1 selects roughly 80k proteins with valid surface data. Appendix A.3 shows low overlap with the test set, but this subset is much smaller than the 40M backbone-only set, and its representativeness is not proven.
  • ad hoc to paper Restricting surface training to proteins containing only C, N, O, S, Se, and H atoms does not bias PRIDE test-set performance.
    Section 3.2 discards proteins with other atom types. If PRIDE test proteins with other atoms are also discarded, or if the remaining set is easier to recover, the reported 61.47% is not comparable to baselines evaluated on the full test set.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DS-ProGen: A Dual-Structure Deep Language Model for Functional Protein Design." pith.science (2026). https://pith.science/paper/JOEB5SB2

@misc{pith2026250512511,
  author       = {Pith},
  title        = {Pith review of: DS-ProGen: A Dual-Structure Deep Language Model for Functional Protein Design},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JOEB5SB2}},
  note         = {Machine review of arXiv:2505.12511}
}
read the original abstract

Inverse Protein Folding (IPF) is a critical subtask in the field of protein design, aiming to engineer amino acid sequences capable of folding correctly into a specified three-dimensional (3D) conformation. Although substantial progress has been achieved in recent years, existing methods generally rely on either backbone coordinates or molecular surface features alone, which restricts their ability to fully capture the complex chemical and geometric constraints necessary for precise sequence prediction. To address this limitation, we present DS-ProGen, a dual-structure deep language model for functional protein design, which integrates both backbone geometry and surface-level representations. By incorporating backbone coordinates as well as surface chemical and geometric descriptors into a next-amino-acid prediction paradigm, DS-ProGen is able to generate functionally relevant and structurally stable sequences while satisfying both global and local conformational constraints. On the PRIDE dataset, DS-ProGen attains the current state-of-the-art recovery rate of 61.47%, demonstrating the synergistic advantage of multi-modal structural encoding in protein design. Furthermore, DS-ProGen excels in predicting interactions with a variety of biological partners, including ligands, ions, and RNA, confirming its robust functional retention capabilities.

Figures

Figures reproduced from arXiv: 2505.12511 by the authors.

Figure 1
Figure 1. Design protein sequences based on protein [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. DS-ProGen integrates backbone and surface structural information to enable functional protein design. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Design cases for DS-ProGen and baselines (Length is 403). A lower [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Structural alignment between ground truth structures (blue) and predicted structures (light brown) folded [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Topology-based splitting strategy for training set construction. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Distribution of maximum sequence similarity, maximum TM-Score, and minimum RMSD between each [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Distribution of RMSD (Å) on PRIDE test set [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Histogram and kernel density estimation (KDE) plot of sequence similarity scores between gen￾erated sequence and groundtruth on PRIDE test set. A.5 Additional Case Studies [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Kernel density estimation plots for structures predicted by AlphaFold3 from model-generated sequences, [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Structural alignment between ground truth structures (blue) and predicted structures (light brown) folded [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Structural alignment between ground truth structures (blue) and predicted structures (light brown) folded [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: Structural alignment between ground truth structures (blue) and predicted structures (light brown) folded [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 28 canonical work pages

  1. [1]

    Ballard, Joshua Bambrick, Sebastian W

    Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvunakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, and 29 others. 2024. https://doi.org/10...

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

  2. [2]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

    arXiv 2023

  3. [3]

    Stepan S Batsanov. 2001. Van der waals radii of elements. Inorganic materials, 37(9):871--885

    work page 2001

  4. [4]

    James F Blinn. 1982. A generalization of algebraic surface drawing. ACM transactions on graphics (TOG), 1(3):235--256

    work page 1982

  5. [5]

    Christiam Camacho, George Coulouris, Vahram Avagyan, Ning Ma, Jason Papadopoulos, Kevin Bealer, and Thomas L Madden. 2009. Blast+: architecture and applications. BMC bioinformatics, 10:1--9

    work page 2009

  6. [6]

    Longxing Cao, Brian Coventry, Inna Goreshnik, Buwei Huang, William Sheffler, Joon Sung Park, Kevin M Jude, Iva Markovi \'c , Rameshwar U Kadam, Koen HG Verschueren, and 1 others. 2022. Design of protein-binding proteins from the target structure alone. Nature, 605(7910):551--560

    work page 2022

  7. [7]

    chq1155. 2024. Pride\_benchmark\_proteindesign. https://github.com/chq1155/PRIDE_Benchmark_ProteinDesign. Accessed: 2025-04-26

    work page 2024

  8. [8]

    Rhiju Das and David Baker. 2008. Macromolecular modeling with rosetta. Annu. Rev. Biochem., 77(1):363--382

    work page 2008

Show all 52 references
  1. [9]

    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, and 1 others. 2022. Robust deep learning--based protein sequence design using proteinmpnn. Science, 378(6615):49--56

    2022

  2. [10]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human...

    2019

  3. [11]

    Yuval Eldar, Michael Lindenbaum, Moshe Porat, and Yehoshua Y Zeevi. 1997. The farthest point strategy for progressive image sampling. IEEE transactions on image processing, 6(9):1305--1315

    1997

  4. [12]

    Jessica Emonts and Johannes Felix Buyel. 2023. An overview of descriptors to capture protein properties--tools and perspectives in the context of qsar modeling. Computational and structural biotechnology journal, 21:3234--3247

    2023

  5. [13]

    Noelia Ferruz, Steffen Schmidt, and Birte H \"o cker. 2022. Protgpt2 is a deep unsupervised language model for protein design. Nature communications, 13(1):4348

    2022

  6. [14]

    Zhangyang Gao, Cheng Tan, Pablo Chac \'o n, and Stan Z Li. 2022. Pifold: Toward effective and efficient protein inverse folding. arXiv preprint arXiv:2209.12643

    2022 arXiv

  7. [15]

    Zhangyang Gao, Cheng Tan, Yijie Zhang, Xingran Chen, Lirong Wu, and Stan Z Li. 2023. Proteininvbench: Benchmarking protein inverse folding on diverse tasks, models, and metrics. Advances in Neural Information Processing Systems, 36:68207--68220

    2023

  8. [16]

    Zhangyang Gao, Cheng Tan, Yijie Zhang, Xingran Chen, Lirong Wu, and Stan Z Li. 2024. Proteininvbench: Benchmarking protein inverse folding on diverse tasks, models, and metrics. Advances in Neural Information Processing Systems, 36

    2024

  9. [17]

    Yu Gu, Robert Tinn, Hao Cheng, Michael Lucas, Naoto Usuyama, Xiaodong Liu, Tristan Naumann, Jianfeng Gao, and Hoifung Poon. 2021. Domain-specific language model pretraining for biomedical natural language processing. ACM Transactions on Computing for Healthcare (HEALTH), 3(1):1--23

    2021

  10. [18]

    Thomas Hayes, Roshan Rao, Halil Akin, Nicholas J Sofroniew, Deniz Oktay, Zeming Lin, Robert Verkuil, Vincent Q Tran, Jonathan Deaton, Marius Wiggert, and 1 others. 2025. Simulating 500 million years of evolution with a language model. Science, page eads0018

    2025

  11. [19]

    Michael Heinzinger, Konstantin Weissenow, Joaquin Gomez Sanchez, Adrian Henkel, Milot Mirdita, Martin Steinegger, and Burkhard Rost. 2024. Bilingual language model for protein sequence and structure. NAR Genomics and Bioinformatics, 6(4):lqae150

    2024

  12. [20]

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

    2022

  13. [21]

    Jo \"e l Janin, Ranjit P Bahadur, and Pinak Chakrabarti. 2008. Protein--protein interaction and quaternary structure. Quarterly reviews of biophysics, 41(2):133--180

    2008

  14. [22]

    Michael Jendrusch, Jan O Korbel, and S Kashif Sadiq. 2021. Alphadesign: A de novo protein design framework based on alphafold. Biorxiv, pages 2021--10

    2021

  15. [23]

    Jiyue Jiang, Pengan Chen, Jiuming Wang, Dongchen He, Ziqin Wei, Liang Hong, Licheng Zong, Sheng Wang, Qinze Yu, Zixian Ma, and 1 others. 2025 a . Benchmarking large language models on multiple tasks in bioinformatics nlp with prompting. arXiv preprint arXiv:2503.04013

    2025 arXiv

  16. [24]

    Jiyue Jiang, Sheng Wang, Qintong Li, Lingpeng Kong, and Chuan Wu. 2023. https://doi.org/10.18653/v1/2023.acl-long.593 A cognitive stimulation dialogue system with multi-source knowledge fusion for elders with cognitive impairment . In Proceedings of the 61st Annual Meeting of ...

    2023 doi

  17. [25]

    Jiyue Jiang, Zikang Wang, Yuheng Shan, Heyan Chai, Jiayi Li, Zixian Ma, Xinrui Zhang, and Yu Li. 2025 b . Biological sequence with language model prompting: A survey. arXiv preprint arXiv:2503.04135

    2025 arXiv

  18. [26]

    Bowen Jing, Stephan Eismann, Patricia Suriana, Raphael JL Townshend, and Ron Dror. 2020. Learning from protein structure with geometric vector perceptrons. arXiv preprint arXiv:2009.01411

    2020 arXiv

  19. [27]

    John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Z \' dek, Anna Potapenko, and 1 others. 2021. Highly accurate protein structure prediction with alphafold. nature, 596(7873):583--589

    2021

  20. [28]

    Hyunbin Kim, Milot Mirdita, and Martin Steinegger. 2023. Foldcomp: a library and format for compressing and indexing large protein structure sets. Bioinformatics, 39(4):btad153

    2023

  21. [29]

    Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Nikita Smetanin, Robert Verkuil, Ori Kabeli, Yaniv Shmueli, and 1 others. 2023. Evolutionary-scale prediction of atomic-level protein structure with a language model. Science, 379(6637):1123--1130

    2023

  22. [30]

    Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, and 1 others. 2024. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437

    2024 arXiv

  23. [31]

    Ali Madani, Ben Krause, Eric R Greene, Subu Subramanian, Benjamin P Mohr, James M Holton, Jose Luis Olmos, Caiming Xiong, Zachary Z Sun, Richard Socher, and 1 others. 2023. Large language models generate functional protein sequences across diverse families. Nature Biotechnolog...

    2023

  24. [32]

    Erik Nijkamp, Jeffrey A Ruffolo, Eli N Weinstein, Nikhil Naik, and Ali Madani. 2023. Progen2: exploring the boundaries of protein language models. Cell systems, 14(11):968--978

    2023

  25. [33]

    Pettersen, Thomas D

    Eric F. Pettersen, Thomas D. Goddard, Conrad C. Huang, Elaine C. Meng, Gregory S. Couch, Tristan I. Croll, John H. Morris, and Thomas E. Ferrin. 2021. https://doi.org/10.1002/pro.3943 Ucsf chimerax: Structure visualization for researchers, educators, and developers . Protein S...

    2021 doi

  26. [34]

    Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever, and 1 others. 2018. Improving language understanding by generative pre-training

    2018

  27. [35]

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, and 1 others. 2019. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9

    2019

  28. [36]

    Dorota Roberts, R Keeling, M Tracka, CF Van Der Walle, S Uddin, Jim Warwicker, and R Curtis. 2015. Specific ion and buffer effects on protein--protein interactions of a monoclonal antibody. Molecular pharmaceutics, 12(1):179--193

    2015

  29. [37]

    Surfpro: Functional protein design based on continuous surface

    Zhenqiao Song, Tinglin Huang, Lei Li, and Wengong Jin. Surfpro: Functional protein design based on continuous surface. In Forty-first International Conference on Machine Learning

  30. [38]

    Freyr Sverrisson, Jean Feydy, Bruno E Correia, and Michael M Bronstein. 2021. Fast end-to-end learning on protein surfaces. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15272--15281

    2021

  31. [39]

    Ilya O Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Andreas Steiner, Daniel Keysers, Jakob Uszkoreit, and 1 others. 2021. Mlp-mixer: An all-mlp architecture for vision. Advances in neural information processing sy...

    2021

  32. [40]

    Michel Van Kempen, Stephanie S Kim, Charlotte Tumescheit, Milot Mirdita, Jeongjae Lee, Cameron LM Gilchrist, Johannes S \"o ding, and Martin Steinegger. 2024. Fast and accurate protein structure search with foldseek. Nature biotechnology, 42(2):243--246

    2024

  33. [41]

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

    2017

  34. [42]

    Zhenyu Wang, Zikang Wang, Jiyue Jiang, Pengan Chen, Xiangyu Shi, and Yu Li. 2025. Large language models in bioinformatics: A survey. arXiv preprint arXiv:2503.04490

    2025

  35. [43]

    Yingce Xia, Peiran Jin, Shufang Xie, Liang He, Chuan Cao, Renqian Luo, Guoqing Liu, Yue Wang, Zequn Liu, Yuan-Jyue Chen, and 1 others. 2025. Naturelm: Deciphering the language of nature for scientific discovery. arXiv preprint arXiv:2502.07527

    2025 arXiv

  36. [44]

    Yijia Xiao, Edward Sun, Yiqiao Jin, Qifan Wang, and Wei Wang. 2024. Proteingpt: Multimodal llm for protein property prediction and structure understanding. arXiv preprint arXiv:2408.11363

    2024 arXiv

  37. [45]

    Mingzhi Yuan, Ao Shen, Kexue Fu, Jiaming Guan, Yingfan Ma, Qin Qiao, and Manning Wang. 2023. Proteinmae: masked autoencoder for protein surface self-supervised learning. Bioinformatics, 39(12):btad724

    2023

  38. [46]

    Kaizhi Yue and Ken A Dill. 1992. Inverse protein folding problem: designing polymer sequences. Proceedings of the National Academy of Sciences, 89(9):4163--4167

    1992

  39. [47]

    Gongbo Zhang, Yanting Li, Renqian Luo, Pipi Hu, Zeru Zhao, Lingbo Li, Guoqing Liu, Zun Wang, Ran Bi, Kaiyuan Gao, and 1 others. 2025. Unigenx: Unified generation of sequence and structure with autoregressive diffusion. arXiv preprint arXiv:2503.06687

    2025 arXiv

  40. [48]

    Yang Zhang and Jeffrey Skolnick. 2004. Scoring function for automated assessment of protein structure template quality. Proteins: Structure, Function, and Bioinformatics, 57(4):702--710

    2004

  41. [49]

    Zhongheng Zhang. 2016. Introduction to machine learning: k-nearest neighbors. Annals of translational medicine, 4(11)

    2016

  42. [50]

    Xinyi Zhou, Guangyong Chen, Junjie Ye, Ercheng Wang, Jun Zhang, Cong Mao, Zhanwei Li, Jianye Hao, Xingxu Huang, Jin Tang, and 1 others. 2023. Prorefiner: an entropy-based refining strategy for inverse protein folding with global graph attention. Nature Communications, 14(1):7434

    2023

  43. [51]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  44. [52]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.