Pith. sign in

REVIEW 4 major objections 6 minor 145 references

A systematic review of 68 studies concludes that generative AI is reshaping bioinformatics, with domain-specialized models consistently outperforming general-purpose language models on biological tasks.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 23:57 UTC pith:EKHBYGMH

load-bearing objection Broad but unreproducible survey of GenAI in bioinformatics; the topical coverage is genuinely wider than prior reviews, but the 'systematic review' claim does not hold up. the 4 major comments →

arxiv 2511.03354 v2 pith:EKHBYGMH submitted 2025-11-05 cs.CL cs.AI

Generative Artificial Intelligence in Bioinformatics: A Systematic Review of Models, Applications, and Methodological Advances

Wasimul Karim , Riasad Alvi , Sayeem Been Zaman , Arefin Ittesafun Abian , Mohaimenul Azam Khan Raiaan , Saddam Mukta , Md Rafi Ur Rashid , Md Rafiqul Islam
show 2 more authors
Yakub Sebastian Sami Azam
This is my paper
classification cs.CL cs.AI
keywords generative AIbioinformaticsprotein language modelssingle-cell genomicsdrug discoverymulti-omics integrationsystematic reviewtransformer models
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper is a systematic review of 68 generative-AI studies in bioinformatics, organized around six research questions covering applications, model effectiveness, domain benefits, task improvements, limitations, and datasets. The authors aim to establish that GenAI—especially domain-specialized architectures—advances bioinformatics by improving accuracy, enabling de novo protein and genome design, integrating multi-omics data, and generating synthetic biological data. They report that specialized models trained with biology-aware tokenization and pretraining generally beat general-purpose LLMs on genomic, proteomic, and single-cell tasks, and that molecular, cellular, and textual datasets collectively support model training and generalization. The review also catalogues persistent limitations such as poor scalability, data bias, limited interpretability, and weak out-of-distribution generalization, and proposes future directions centered on modular, tool-grounded, and biologically informed systems. A sympathetic reader would take this as a credible map of where GenAI is succeeding in computational biology and where the evidence is still thin.

Core claim

The central claim is that generative AI has moved from a niche technique to a broadly applicable tool in bioinformatics, with the review's evidence showing that specialized model architectures generally outperform general-purpose models on biological sequence and structure tasks. Across six research questions, the authors find that GenAI supports sequence analysis, molecular design, and integrative data modeling; that domain-specific pretraining and context-aware tokenization drive performance gains; that protein structure prediction, functional annotation, and synthetic data generation have advanced substantially; and that a diverse set of molecular, cellular, and textual datasets enables t

What carries the argument

The analytical machinery is the six-research-question framework, which organizes the reviewed literature into applications, model architectures, domain benefits, task-level improvements, limitations, and datasets. Within that framework, the load-bearing comparison is between domain-specialized models—such as protein language models and k-mer-aware genomic transformers—and general-purpose LLMs. The mechanism that carries the argument is domain-specific pretraining: when tokenization and training data are tailored to biological sequences, models capture contextual and structural patterns that general-purpose models miss, and this specialization, rather than raw parameter count, is credited for

Load-bearing premise

The review's central conclusions rest on the assumption that the 68 selected studies are representative of the field and that the performance numbers quoted from them are trustworthy and comparable, despite differences in benchmarks, evaluation protocols, and data splits.

What would settle it

A direct head-to-head evaluation would settle the main claim: take a single biological task (e.g., promoter prediction or protein function annotation), train or fine-tune a general-purpose LLM and a domain-specialized model under matched data, compute, and evaluation settings, and compare performance; if the general-purpose model matches or beats the specialized model across several tasks, the review's specialization conclusion weakens substantially.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Researchers choosing models for biological sequence tasks should prefer domain-specialized pretrained models over general-purpose LLMs when such models are available, because the reviewed evidence shows consistent gains from biologically informed tokenization and pretraining.
  • Reported improvements in protein structure prediction, mutation-effect prediction, and de novo protein generation suggest that alignment-free, embedding-based methods can scale to metagenomic and under-studied organisms where traditional multiple-sequence-alignment approaches are impractical.
  • The dataset analysis implies that high-quality training and evaluation depend on combining molecular, cellular, and textual resources; investigators building GenAI systems for bioinformatics should expect multimodal data to improve generalization.
  • The catalogue of limitations—scalability cost, data bias, hallucination, and weak out-of-distribution performance—implies that deployment in clinical or high-stakes settings requires uncertainty quantification, interpretability tools, and grounding in verifiable bioinformatics software.
  • The review's future-directions argument implies a shift from monolithic models to modular systems in which a generalist LLM orchestrates specialized tools, reducing hallucination and increasing reproducibility in complex workflows.

Where Pith is reading between the lines

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

  • If the specialization advantage holds as a general pattern, a testable extension would be that biology-aware tokenization and pretraining will also benefit emerging modalities such as spatial omics, metagenomics, and RNA engineering, where current evidence is thinner.
  • Because the performance numbers in the review come from heterogeneous benchmarks without normalization, an editorial inference is that a unified, standardized evaluation suite would likely shrink some reported gaps between specialized and general-purpose models, even if it did not erase them.
  • The review's emphasis on modular, tool-grounded systems suggests a concrete next step: measuring whether LLM agents that delegate to verified bioinformatics tools reduce hallucination rates and improve reproducibility relative to end-to-end generative models; this is an implication the authors gesture toward but do not test.
  • Given the rapid growth of publications after 2023, one inference is that the field may consolidate around a small number of foundation models shared across many tasks, making dataset quality and benchmark standardization even more critical than algorithmic novelty.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. This manuscript claims to be a PRISMA-guided systematic review of generative artificial intelligence in bioinformatics, synthesizing 68 papers published between 2021 and 2025 to answer six research questions about applications, model architectures, domain benefits, task-level improvements, limitations, and datasets. The headline conclusions are that GenAI is broadly useful across genomics, proteomics, transcriptomics, and drug discovery; that domain-specific architectures generally outperform general-purpose LLMs; and that molecular, cellular, and textual datasets collectively support model training and generalization. The paper is primarily a narrative synthesis supported by descriptive tables and a taxonomic framework.

Significance. If the evidence base were auditable, the review would fill a useful niche by comparing five prior reviews and extending coverage to multi-agent systems, conversational interfaces, and workflow automation. The organization around six RQs and eleven topics is clear, and the dataset inventory in Table 6 could become a practical resource after correction. However, the central claims — especially the abstract's assertion that the review 'demonstrates the growing potential of GenAI' and RQ2's conclusion that 'specialized model architectures generally outperform general-purpose models' — depend entirely on a selection of 68 papers and on performance figures quoted from heterogeneous sources. Because the selection process is not reproducible and the performance numbers are not comparable, the evidentiary value of the current manuscript is closer to that of an expert narrative than a systematic review. No reproducible search script, screening log, or analysis code is provided; these absences are load-bearing for the review's stated methodology.

major comments (4)
  1. [§3.1, §3.3] The search strategy is not reproducible. §3.1 lists only broad terms such as 'GenAI in bioinformatics' and 'Protein language models' with no exact query strings, database-specific syntax, search date, or number of records retrieved per source. §3.3 reports only the final count of 68 papers and their venue distribution; no PRISMA flow diagram, screening log, or excluded-study list is provided. Since the abstract claims adherence to PRISMA, and since every RQ answer in §4 rests on this selection, the audit trail must be supplied (as supplementary material) or the systematic-review claim must be withdrawn.
  2. [§4.2–§4.4, Table 3] Performance comparisons are drawn from heterogeneous benchmarks without normalization or matched baselines. Table 3 mixes DNABERT accuracy (91.8–91.9%), TrustAffinity RMSE (0.384), MegaMolBART AUC (0.981), scGPT/scBERT accuracy (84%), and BioGPT PubMedQA accuracy (78.2%). These numbers come from different tasks, datasets, and splits, so the claim that specialized models 'generally outperform' general-purpose models is a synthesis across incomparable figures. The authors should either restrict the claim to within-study controlled comparisons or mark each number with task, dataset, split, and whether it is source-reported or independently re-evaluated.
  3. [§4.2.1 vs §4.4.3] There is an internal tension about scGPT. §4.2.1 cites [58] to state that 'in zero-shot reconstruction settings, scGPT does not consistently exceed trivial predictors.' Later, §4.4.3 states that scGPT shows 'SOTA accuracy in cell type annotation of up to 98% and macro F1-scores consistently higher than baselines such as scBERT, GEARS, and scGLUE.' These two statements need to be reconciled, or the conditions under which each holds must be stated explicitly, otherwise the review sends contradictory messages about a model it repeatedly highlights.
  4. [Table 4, Table 6, and reference list] Several citation and reference errors undermine trust in the curation. Table 4 lists 'Aliro [48]' but reference [48] is the scGPT paper; Table 4 also lists 'sciCAN [54]' while sciCAN is [53] and scDREAMER is [54]; Table 6 lists 'DisGenNet [81]' which points to a 2020 EMNLP summarization paper rather than the DisGeNET database; Table 6 lists 'UniPort Gene-Protein pair [110]' for a paper describing ChEBI ligand annotation; and Table 6's 'PMKB-CV Disease-gene associations [87]' points to the LORE paper. These errors affect dataset and method attribution. A systematic audit of all references and table entries is required.
minor comments (6)
  1. [§1 Introduction] 'present fearful challenges' should be 'present daunting challenges'; likewise, 'reconstructing computational biology' and 'gene controlling element' are infelicitous. A careful language pass is needed.
  2. [§4.2.2] Typo: 'ourperforming' should be 'outperforming'.
  3. [§4.4.1, §4.4.2] 'remainders' and 'inter-remains' should be 'residues' and 'inter-residue' throughout.
  4. [§4.5.1] 'transistor architectures' should be 'transformer architectures'.
  5. [Table 6] Inconsistent naming: 'UniPort' should be 'UniProt'; 'DisGenNet' should be 'DisGeNET'; 'Gene expresion' should be 'Gene expression'.
  6. [General] Some sentences in §4.3 and §5 are run-ons with repeated references to 'previous sections' and 'aforementioned'; a structural edit would improve readability.

Circularity Check

0 steps flagged

No circularity found: the review synthesizes external primary studies and does not fit its own inputs or rely on a self-citation chain.

full rationale

This paper is a systematic literature review, not a derivation chain with fitted equations. Its central claims, such as RQ2's conclusion that 'domain-specific models... consistently outperform general-purpose LLMs on various sets of bioinformatics tasks' (Section 4.2.1), are supported by performance numbers quoted from independent primary studies (DNABERT, ESM-1v, scGPT, TrustAffinity, etc.). There is no equation whose predicted quantity is identical to a fitted input by construction, no fitted parameter is renamed as a prediction, and no uniqueness theorem or ansatz is imported through the authors' own prior work. The authors do not appear to rely on load-bearing self-citations; the cited prior reviews ([13, 6, 14, 15, 16]) are used for context and gap analysis, not to force the review's conclusions. The main integrity concerns are evidentiary rather than circular: the PRISMA-style selection is not fully auditable (Section 3.3 reports only the final count of 68 articles without a flow diagram or excluded-study list), and some internal citations are inconsistent (e.g., sciCAN cited as [54] in Table 4; DisGeNET reference [81] points to a summarization paper). These issues affect reproducibility and soundness, but they do not make the review's claims equivalent to its inputs by construction. Under the specified circularity criteria, the honest finding is no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

No free parameters or invented entities are relevant because the paper is a literature review. The central claims depend on three unverified domain assumptions: unrepresentative selection, comparability of heterogeneous metrics, and correctness of the citation mapping.

axioms (3)
  • domain assumption The 68 selected articles are representative of the GenAI bioinformatics literature.
    Section 3.2-3.3 gives inclusion/exclusion criteria and a final count, but no PRISMA flow, screening logs, or excluded-study list; all RQ answers assume this set is complete and unbiased.
  • domain assumption Performance numbers from different papers are directly comparable across tasks.
    Sections 4.2-4.4 quote accuracy, AUC, Spearman, F1 from different datasets and compare them directly; no benchmark harmonization or effect-size synthesis is provided.
  • domain assumption Cited references correctly support the attributed claims.
    The paper depends on citations such as DisGeNET [81] and Aliro [48]; several are mismatched, so this assumption is violated in places.

pith-pipeline@v1.3.0-alltime-deepseek · 43150 in / 11694 out tokens · 105095 ms · 2026-08-03T23:57:03.149221+00:00 · methodology

0 comments
read the original abstract

Generative artificial intelligence (GenAI) is transforming bioinformatics by advancing genomics, proteomics, transcriptomics, structural biology, and drug discovery. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework, this review addresses six research questions to evaluate influential GenAI strategies in terms of methodological innovation, predictive performance, specialization, limitations, and data use. RQ1 shows that GenAI supports sequence analysis, molecular design, and integrative data modelling, often outperforming traditional methods through improved pattern recognition and generation. RQ2 finds that specialized architectures generally outperform general-purpose models because of domain-specific pretraining and context-aware design. RQ3 identifies benefits in molecular analysis and biological data integration, including improved accuracy and reduced analytical error. RQ4 reports advances in structural modelling, functional prediction, and synthetic data generation, supported by established benchmarks. RQ5 highlights key limitations, including poor scalability, data bias, and restricted generalizability, and recommends stronger evaluation and biologically grounded modelling. RQ6 shows that molecular datasets, including UniProtKB and ProteinNet12, cellular datasets, including CELLxGENE and GTEx, and textual resources, including PubMedQA and OMIM, support model training and generalization. Overall, this review demonstrates the growing potential of GenAI to advance computational biology through more accurate, specialized, and integrative bioinformatics analysis.

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

145 extracted references · 14 linked inside Pith

  1. [1]

    Arpita Das, Manojit Bhattacharya, Ali Saber Abdel- hameed, Sang-Soo Lee, and Chiranjib Chakraborty. Generative artificial intelligence, integrative bioinfor- matics, and single-cell analysis reveal alzheimer’s ge- netic and immune landscape.Molecular Therapy Nu- cleic Acids, 36(2), 2025

    2025

  2. [2]

    Ma- chine learning applications in genetics and genomics

    Maxwell W Libbrecht and William Stafford Noble. Ma- chine learning applications in genetics and genomics. Nature Reviews Genetics, 16(6):321–332, 2015

    2015

  3. [3]

    Using rule-based machine learn- ing for candidate disease gene prioritization and sample classification of cancer gene expression data.PloS one, 7(7):e39932, 2012

    Enrico Glaab, Jaume Bacardit, Jonathan M Garibaldi, and Natalio Krasnogor. Using rule-based machine learn- ing for candidate disease gene prioritization and sample classification of cancer gene expression data.PloS one, 7(7):e39932, 2012

    2012

  4. [4]

    Deep learning for computational biology.Molecular systems biology, 12(7):878, 2016

    ChristofAngermueller, TanelPärnamaa, LeopoldParts, and Oliver Stegle. Deep learning for computational biology.Molecular systems biology, 12(7):878, 2016

    2016

  5. [5]

    Applications of transformer- based language models in bioinformatics: a survey

    Shuang Zhang, Rui Fan, Yuti Liu, Shuang Chen, Qiao Liu, and Wanwen Zeng. Applications of transformer- based language models in bioinformatics: a survey. Bioinformatics Advances, 3(1):vbad001, 2023

    2023

  6. [6]

    Foundation models in bioinformatics.National science review, 12(4):nwaf028, 2025

    Fei Guo, Renchu Guan, Yaohang Li, Qi Liu, Xiaowo Wang, Can Yang, and Jianxin Wang. Foundation models in bioinformatics.National science review, 12(4):nwaf028, 2025

    2025

  7. [7]

    Large language models for bioinformatics

    WeiRuan, YanjunLyu, JingZhang, JiazhangCai, Peng Shu, Yang Ge, Yao Lu, Shang Gao, Yue Wang, Peilong Wang, et al. Large language models for bioinformatics. Quantitative Biology, 14(1):e70014, 2026

    2026

  8. [8]

    Generative models for protein se- quence modeling: recent advances and future directions

    Mehrsa Mardikoraem, Zirui Wang, Nathaniel Pascual, and Daniel Woldring. Generative models for protein se- quence modeling: recent advances and future directions. Briefings in Bioinformatics, 24(6):bbad358, 2023

    2023

  9. [9]

    Language models enable zero-shot prediction of the effects of mutations on pro- tein function.Advances in neural information process- ing systems, 34:29287–29303, 2021

    Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu, and Alex Rives. Language models enable zero-shot prediction of the effects of mutations on pro- tein function.Advances in neural information process- ing systems, 34:29287–29303, 2021

    2021

  10. [10]

    Gen- erative design of de novo proteins based on secondary- structure constraints using an attention-based diffusion model.Chem, 9(7):1828–1849, 2023

    Bo Ni, David L Kaplan, and Markus J Buehler. Gen- erative design of de novo proteins based on secondary- structure constraints using an attention-based diffusion model.Chem, 9(7):1828–1849, 2023

    2023

  11. [11]

    Qust-llm: Integrating large language models for comprehensive spatial transcriptomics anal- ysis.arXiv preprint arXiv:2406.14307, 2024

    Chao Hui Huang. Qust-llm: Integrating large language models for comprehensive spatial transcriptomics anal- ysis.arXiv preprint arXiv:2406.14307, 2024. 33

    Pith/arXiv arXiv 2024

  12. [12]

    Generative ai for synthetic data across multiple medical modalities: A systematic review of recent developments and chal- lenges.Computers in biology and medicine, 189:109834, 2025

    Mahmoud Ibrahim, Yasmina Al Khalil, Sina Amirrajab, Chang Sun, Marcel Breeuwer, Josien Pluim, Bart Elen, Gökhan Ertaylan, and Michel Dumontier. Generative ai for synthetic data across multiple medical modalities: A systematic review of recent developments and chal- lenges.Computers in biology and medicine, 189:109834, 2025

    2025

  13. [13]

    Progress and opportunities of foundation models in bioinformatics.Briefings in Bioinformatics, 25(6):bbae548, 2024

    Qing Li, Zhihang Hu, Yixuan Wang, Lei Li, Yimin Fan, Irwin King, Gengjie Jia, Sheng Wang, Le Song, and Yu Li. Progress and opportunities of foundation models in bioinformatics.Briefings in Bioinformatics, 25(6):bbae548, 2024

    2024

  14. [14]

    Large lan- guage models and their applications in bioinformatics

    Oluwafemi A Sarumi and Dominik Heider. Large lan- guage models and their applications in bioinformatics. Computational and Structural Biotechnology Journal, 23:3498–3505, 2024

    2024

  15. [15]

    Muhammad Nabeel Asim, Muhammad Ali Ibrahim, AroojZaib, andAndreasDengel. Dnasequenceanalysis landscape: a comprehensive review of dna sequence analysis task types, databases, datasets, word embed- ding methods, and language models.Frontiers in Medicine, 12:1503229, 2025

    2025

  16. [16]

    Protein sequence analysis land- scape: A systematic review of task types, databases, datasets, word embeddings methods, and language models.Database, 2025:baaf027, 2025

    Muhammad Nabeel Asim, Tayyaba Asif, Faiza Hassan, and Andreas Dengel. Protein sequence analysis land- scape: A systematic review of task types, databases, datasets, word embeddings methods, and language models.Database, 2025:baaf027, 2025

    2025

  17. [17]

    Dnabert: pre-trained bidirectional en- coder representations from transformers model for dna- language in genome.Bioinformatics, 37(15):2112–2120, 2021

    Yanrong Ji, Zhihan Zhou, Han Liu, and Ramana V Davuluri. Dnabert: pre-trained bidirectional en- coder representations from transformers model for dna- language in genome.Bioinformatics, 37(15):2112–2120, 2021

    2021

  18. [18]

    Dna language model grover learns se- quence context in the human genome.Nature Machine Intelligence, 6(8):911–923, 2024

    Melissa Sanabria, Jonas Hirsch, Pierre M Joubert, and Anna R Poetsch. Dna language model grover learns se- quence context in the human genome.Nature Machine Intelligence, 6(8):911–923, 2024

    2024

  19. [19]

    Dnabert-2: Efficientfoun- dation model and benchmark for multi-species genome

    Zhihan Zhou, Yanrong Ji, Weijian Li, Pratik Dutta, Ra- mana Davuluri, andHan Liu. Dnabert-2: Efficientfoun- dation model and benchmark for multi-species genome. arXiv preprint arXiv:2306.15006, 2023

    Pith/arXiv arXiv 2023

  20. [20]

    Nucleotide transformer: building and eval- uating robust foundation models for human genomics

    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 Sirelkha- tim, et al. Nucleotide transformer: building and eval- uating robust foundation models for human genomics. Nature Methods, 22(2):287–297, 2025

    2025

  21. [21]

    Yumeng Zhang, Yangming Zhang, Yi Xiong, Hui Wang, Zixin Deng, Jiangning Song, and Hong-Yu Ou. T4sefinder: a bioinformatics tool for genome-scale pre- diction of bacterial type iv secreted effectors using pre- trained protein language model.Briefings in Bioinfor- matics, 23(1):bbab420, 2022

    2022

  22. [22]

    Gp-gpt: Large language model for gene-phenotype mapping.arXiv preprint arXiv:2409.09825, 2024

    Yanjun Lyu, Zihao Wu, Lu Zhang, Jing Zhang, Yi- wei Li, Wei Ruan, Zhengliang Liu, Xiaowei Yu, Chao Cao, Tong Chen, et al. Gp-gpt: Large language model for gene-phenotype mapping.arXiv preprint arXiv:2409.09825, 2024

    arXiv 2024

  23. [23]

    Enhancing gene set overrepresentation analysis with large language models.Bioinformatics Advances, 5(1):vbaf054, 2025

    Jiqing Zhu, Rebecca Y Wang, Xiaoting Wang, Ricardo Azevedo, Alexander Moreno, Julia A Kuhn, and Zia Khan. Enhancing gene set overrepresentation analysis with large language models.Bioinformatics Advances, 5(1):vbaf054, 2025

    2025

  24. [24]

    Denis Sidorenko, Stefan Pushkov, Akhmed Sakip, Ge- offrey Ho Duen Leung, Sarah Wing Yan Lok, Anatoly Urban, Diana Zagirova, Alexander Veviorskiy, Nina Ti- honova, Aleksandr Kalashnikov, et al. Precious2gpt: the combination of multiomics pretrained transformer and conditional diffusion for artificial multi-omics multi- species multi-tissue sample generatio...

    2024

  25. [25]

    Bioagents: Democratizing bioinformatics analysis with multi-agent systems.arXiv preprint arXiv:2501.06314, 2025

    NikitaMehandru, AmandaKHall, OlesyaMelnichenko, Yulia Dubinina, Daniel Tsirulnikov, David Bamman, Ahmed Alaa, Scott Saponas, and Venkat S Malladi. Bioagents: Democratizing bioinformatics analysis with multi-agent systems.arXiv preprint arXiv:2501.06314, 2025

    Pith/arXiv arXiv 2025

  26. [26]

    Large lan- guage model consensus substantially improves the cell type annotation accuracy for scrna-seq data.bioRxiv, pages 2025–04, 2025

    Chen Yang, Xianyang Zhang, and Jun Chen. Large lan- guage model consensus substantially improves the cell type annotation accuracy for scrna-seq data.bioRxiv, pages 2025–04, 2025

    2025

  27. [27]

    ‘bingo’—a large language model- and graph neural network-based workflow for the pre- diction of essential genes from protein data.Briefings in Bioinformatics, 25(1):bbad472, 2024

    Jiani Ma, Jiangning Song, Neil D Young, Bill CH Chang, Pasi K Korhonen, Tulio L Campos, Hui Liu, and Robin B Gasser. ‘bingo’—a large language model- and graph neural network-based workflow for the pre- diction of essential genes from protein data.Briefings in Bioinformatics, 25(1):bbad472, 2024

    2024

  28. [28]

    Protein language-model embeddings for fast, accurate, and alignment-free protein structure prediction.Structure, 30(8):1169–1177, 2022

    Konstantin Weissenow, Michael Heinzinger, and Burkhard Rost. Protein language-model embeddings for fast, accurate, and alignment-free protein structure prediction.Structure, 30(8):1169–1177, 2022

    2022

  29. [29]

    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

  30. [30]

    Jeffrey Skolnick, Mu Gao, Hongyi Zhou, and Suresh Singh. Alphafold 2: why it works and its implications 34 for understanding the relationships of protein sequence, structure, and function.Journal of chemical informa- tion and modeling, 61(10):4827–4831, 2021

    2021

  31. [31]

    Yuhuan Liu, Xueying Wang, Haitian Zhong, Jixiu Zhai, Xiaojuan Gong, and Tianchi Lu. Phosf3c: a fea- ture fusion architecture with fine-tuned protein lan- guage model and conformer for prediction of gen- eral phosphorylation site.Briefings in Bioinformatics, 26(3):bbaf242, 2025

    2025

  32. [32]

    plm- blast: distant homology detection based on direct com- parison of sequence representations from protein lan- guage models.Bioinformatics, 39(10):btad579, 2023

    Kamil Kaminski, Jan Ludwiczak, Kamil Pawlicki, Vikram Alva, and Stanislaw Dunin-Horkawicz. plm- blast: distant homology detection based on direct com- parison of sequence representations from protein lan- guage models.Bioinformatics, 39(10):btad579, 2023

    2023

  33. [33]

    Improvements in viral gene annotation using large language models and soft alignments.BMC bioinformatics, 25(1):165, 2024

    William L Harrigan, Barbra D Ferrell, K Eric Wom- mack, Shawn W Polson, Zachary D Schreiber, and Mahdi Belcaid. Improvements in viral gene annotation using large language models and soft alignments.BMC bioinformatics, 25(1):165, 2024

    2024

  34. [34]

    Trustaffinity: accurate, reliable and scalable out-of- distribution protein-ligand binding affinity prediction using trustworthy deep learning.bioRxiv, 2024

    Amitesh Badkul, Li Xie, Shuo Zhang, and Lei Xie. Trustaffinity: accurate, reliable and scalable out-of- distribution protein-ligand binding affinity prediction using trustworthy deep learning.bioRxiv, 2024

    2024

  35. [35]

    Protchatgpt: Towards understanding proteins with large language models.arXiv preprint arXiv:2402.09649, 2024

    Chao Wang, Hehe Fan, Ruijie Quan, and Yi Yang. Protchatgpt: Towards understanding proteins with large language models.arXiv preprint arXiv:2402.09649, 2024

    Pith/arXiv arXiv 2024

  36. [36]

    Drbioright 2.0: an llm- powered bioinformatics chatbot for large-scale cancer functional proteomics analysis.Nature communications, 16(1):2256, 2025

    Wei Liu, Jun Li, Yitao Tang, Yining Zhao, Chaozhong Liu, Meiyi Song, Zhenlin Ju, Shwetha V Kumar, Yil- ing Lu, Rehan Akbani, et al. Drbioright 2.0: an llm- powered bioinformatics chatbot for large-scale cancer functional proteomics analysis.Nature communications, 16(1):2256, 2025

    2025

  37. [37]

    Prot2chat: protein largelanguagemodelwithearlyfusionoftext, sequence, and structure.Bioinformatics, 41(8):btaf396, 2025

    Zhicong Wang, Zicheng Ma, Ziqiang Cao, Changlong Zhou, Jun Zhang, and Yi Qin Gao. Prot2chat: protein largelanguagemodelwithearlyfusionoftext, sequence, and structure.Bioinformatics, 41(8):btaf396, 2025

    2025

  38. [38]

    Prot- gpt2 is a deep unsupervised language model for protein design.Nature communications, 13(1):4348, 2022

    Noelia Ferruz, Steffen Schmidt, and Birte Höcker. Prot- gpt2 is a deep unsupervised language model for protein design.Nature communications, 13(1):4348, 2022

    2022

  39. [39]

    Progen: Lan- guage modeling for protein generation.arXiv preprint arXiv:2004.03497, 2020

    Ali Madani, Bryan McCann, Nikhil Naik, Nitish Shirish Keskar, Namrata Anand, Raphael R Eguchi, Po- Ssu Huang, and Richard Socher. Progen: Lan- guage modeling for protein generation.arXiv preprint arXiv:2004.03497, 2020

    Pith/arXiv arXiv 2004

  40. [40]

    Progen2: exploring the boundaries of protein language models.Cell systems, 14(11):968–978, 2023

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

    2023

  41. [41]

    Large language models generate func- tional protein sequences across diverse families.Nature biotechnology, 41(8):1099–1106, 2023

    Ali Madani, Ben Krause, Eric R Greene, Subu Subra- manian, Benjamin P Mohr, James M Holton, Jose Luis Olmos Jr, Caiming Xiong, Zachary Z Sun, Richard Socher, et al. Large language models generate func- tional protein sequences across diverse families.Nature biotechnology, 41(8):1099–1106, 2023

    2023

  42. [42]

    xtrimopglm: unified 100b-scale pre-trained transformer for deciphering the language of protein

    Bo Chen, Xingyi Cheng, Pan Li, Yangli-ao Geng, Jing Gong, Shen Li, Zhilei Bei, Xu Tan, Boyan Wang, Xin Zeng, et al. xtrimopglm: unified 100b-scale pre-trained transformer for deciphering the language of protein. arXiv preprint arXiv:2401.06199, 2024

    Pith/arXiv arXiv 2024

  43. [43]

    Deep generative models of protein structure uncover distant relationships across a contin- uous fold space.Nature Communications, 15(1):8094, 2024

    Eli J Draizen, Stella Veretnik, Cameron Mura, and Philip E Bourne. Deep generative models of protein structure uncover distant relationships across a contin- uous fold space.Nature Communications, 15(1):8094, 2024

    2024

  44. [44]

    Exploiting pretrained biochemical language models for targeted drug design

    Gökçe Uludoğan, Elif Ozkirimli, Kutlu O Ulgen, Nil- gün Karalı, and Arzucan Özgür. Exploiting pretrained biochemical language models for targeted drug design. Bioinformatics, 38(Supplement_2):ii155–ii161, 2022

    2022

  45. [45]

    Drugassist: A large language model for molecule optimization.Briefings in Bioinformatics, 26(1):bbae693, 2025

    Geyan Ye, Xibao Cai, Houtim Lai, Xing Wang, Jun- hong Huang, Longyue Wang, Wei Liu, and Xiangxi- ang Zeng. Drugassist: A large language model for molecule optimization.Briefings in Bioinformatics, 26(1):bbae693, 2025

    2025

  46. [46]

    Thu, Dimitar Dobchev, Akshay Balsubramani, Ricardo ND Santos, Lorenzo Kogler- Anele, Alejandro Corrochano-Navarro, et al

    Saeed Moayedpour, Jonathan Broadbent, Saleh Riahi, Michael Bailey, Hoa V. Thu, Dimitar Dobchev, Akshay Balsubramani, Ricardo ND Santos, Lorenzo Kogler- Anele, Alejandro Corrochano-Navarro, et al. Represen- tations of lipid nanoparticles using large language mod- els for transfection efficiency prediction.Bioinformatics, 40(7):btae342, 2024

    2024

  47. [47]

    Top-dti: Integrat- ing topological deep learning and large language models for drug target interaction prediction.bioRxiv, 2025

    Muhammed Talo and Serdar Bozdag. Top-dti: Integrat- ing topological deep learning and large language models for drug target interaction prediction.bioRxiv, 2025

    2025

  48. [48]

    scgpt: toward building a foundation model for single-cell multi-omics using generative ai.Nature Methods, 21(8):1470–1480, 2024

    Haotian Cui, Chloe Wang, Hassaan Maan, Kuan Pang, Fengning Luo, Nan Duan, and Bo Wang. scgpt: toward building a foundation model for single-cell multi-omics using generative ai.Nature Methods, 21(8):1470–1480, 2024

    2024

  49. [49]

    Olaf: An open life science analysis framework for conversational bioinformatics powered by large lan- guage models.arXiv preprint arXiv:2504.03976, 2025

    Dylan Riffle, Nima Shirooni, Cody He, Manush Murali, Sovit Nayak, Rishikumar Gopalan, and Diego Gonzalez Lopez. Olaf: An open life science analysis framework for conversational bioinformatics powered by large lan- guage models.arXiv preprint arXiv:2504.03976, 2025

    Pith/arXiv arXiv 2025

  50. [50]

    Language model powered digital biology with brad.arXiv preprint arXiv:2409.02864, 2024

    Joshua Pickard, Ram Prakash, Marc Andrew Choi, Natalie Oliven, Cooper Stansbury, Jillian Cwycyshyn, Alex Gorodetsky, Alvaro Velasquez, and Indika Ra- japakse. Language model powered digital biology with brad.arXiv preprint arXiv:2409.02864, 2024. 35

    Pith/arXiv arXiv 2024

  51. [51]

    Pyevocell: an llm-augmented single- cell trajectory analysis dashboard.Bioinformatics, 41(4):btaf158, 2025

    SachinMathur, MathieuBeauvais, ArnauGiribet, Nico- las Aragon Barrero, Chaorui-Tom Zhang, Towsif Rah- man, Seqian Wang, Jeremy Huang, Nima Nouri, Andre Kurlovs, et al. Pyevocell: an llm-augmented single- cell trajectory analysis dashboard.Bioinformatics, 41(4):btaf158, 2025

    2025

  52. [52]

    Integrating spatial and single-cell transcriptomics data using deep generative models with spatialscope.Nature Communications, 14(1):7848, 2023

    Xiaomeng Wan, Jiashun Xiao, Sindy Sing Ting Tam, Mingxuan Cai, Ryohichi Sugimura, Yang Wang, Xiang Wan, Zhixiang Lin, Angela Ruohao Wu, and Can Yang. Integrating spatial and single-cell transcriptomics data using deep generative models with spatialscope.Nature Communications, 14(1):7848, 2023

    2023

  53. [53]

    scican: single-cell chromatin accessibility and gene ex- pression data integration via cycle-consistent adversar- ial network.NPJ Systems Biology and Applications, 8(1):33, 2022

    Yang Xu, Edmon Begoli, and Rachel Patton McCord. scican: single-cell chromatin accessibility and gene ex- pression data integration via cycle-consistent adversar- ial network.NPJ Systems Biology and Applications, 8(1):33, 2022

    2022

  54. [54]

    scdreamer for atlas-level integration of single-cell datasets using deep generative model paired with ad- versarial classifier.Nature Communications, 14(1):7781, 2023

    Ajita Shree, Musale Krushna Pavan, and Hamim Za- far. scdreamer for atlas-level integration of single-cell datasets using deep generative model paired with ad- versarial classifier.Nature Communications, 14(1):7781, 2023

    2023

  55. [55]

    Evaluating largelanguagemodelsforannotatingproteins.Briefings in Bioinformatics, 25(3):bbae177, 2024

    Rosario Vitale, Leandro A Bugnon, Emilio Luis Fenoy, Diego H Milone, and Georgina Stegmayer. Evaluating largelanguagemodelsforannotatingproteins.Briefings in Bioinformatics, 25(3):bbae177, 2024

    2024

  56. [56]

    scbert as a large-scale pretrained deep language model for cell type annotation of single-cell rna-seq data.Na- ture Machine Intelligence, 4(10):852–866, 2022

    Fan Yang, Wenchuan Wang, Fang Wang, Yuan Fang, Duyu Tang, Junzhou Huang, Hui Lu, and Jianhua Yao. scbert as a large-scale pretrained deep language model for cell type annotation of single-cell rna-seq data.Na- ture Machine Intelligence, 4(10):852–866, 2022

    2022

  57. [57]

    Gears: Predicting transcriptional outcomes of novel multi-gene perturbations.BioRxiv, pages 2022–07, 2022

    Yusuf Roohani, Kexin Huang, and Jure Leskovec. Gears: Predicting transcriptional outcomes of novel multi-gene perturbations.BioRxiv, pages 2022–07, 2022

    2022

  58. [58]

    Zero-shot evaluation reveals limita- tions of single-cell foundation models.Genome Biology, 26(1):101, 2025

    Kasia Z Kedzierska, Lorin Crawford, Ava P Amini, and Alex X Lu. Zero-shot evaluation reveals limita- tions of single-cell foundation models.Genome Biology, 26(1):101, 2025

    2025

  59. [59]

    Assessing the state of the art in biomedical relation extraction: overview of the biocreative v chemical-disease relation (cdr) task.Database, 2016:baw032, 2016

    Chih-Hsuan Wei, Yifan Peng, Robert Leaman, Al- lan Peter Davis, Carolyn J Mattingly, Jiao Li, Thomas C Wiegers, and Zhiyong Lu. Assessing the state of the art in biomedical relation extraction: overview of the biocreative v chemical-disease relation (cdr) task.Database, 2016:baw032, 2016

    2016

  60. [60]

    Discover- ing drug–target interaction knowledge from biomedical literature.Bioinformatics, 38(22):5100–5107, 2022

    Yutai Hou, Yingce Xia, Lijun Wu, Shufang Xie, Yang Fan, Jinhua Zhu, Tao Qin, and Tie-Yan Liu. Discover- ing drug–target interaction knowledge from biomedical literature.Bioinformatics, 38(22):5100–5107, 2022

    2022

  61. [61]

    The ddi corpus: An annotated corpus with pharmacological substances and drug–drug interactions.Journal of biomedical informat- ics, 46(5):914–920, 2013

    María Herrero-Zazo, Isabel Segura-Bedmar, Paloma Martínez, and Thierry Declerck. The ddi corpus: An annotated corpus with pharmacological substances and drug–drug interactions.Journal of biomedical informat- ics, 46(5):914–920, 2013

    2013

  62. [62]

    Pubmedqa: A dataset for biomedical research question answering.arXiv preprint arXiv:1909.06146, 2019

    Qiao Jin, Bhuwan Dhingra, Zhengping Liu, William W Cohen, and Xinghua Lu. Pubmedqa: A dataset for biomedical research question answering.arXiv preprint arXiv:1909.06146, 2019

    Pith/arXiv arXiv 1909

  63. [63]

    Biogpt: generative pre-trained transformer for biomedical text generation and mining.Briefings in bioinformatics, 23(6):bbac409, 2022

    Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon, and Tie-Yan Liu. Biogpt: generative pre-trained transformer for biomedical text generation and mining.Briefings in bioinformatics, 23(6):bbac409, 2022

    2022

  64. [64]

    Biomedlm: A 2.7 b parameter language model trained on biomedical text.arXiv preprint arXiv:2403.18421, 2024

    Elliot Bolton, Abhinav Venigalla, Michihiro Yasunaga, David Hall, Betty Xiong, Tony Lee, Roxana Daneshjou, Jonathan Frankle, Percy Liang, Michael Carbin, et al. Biomedlm: A 2.7 b parameter language model trained on biomedical text.arXiv preprint arXiv:2403.18421, 2024

    Pith/arXiv arXiv 2024

  65. [65]

    A decision-aid model for pre- dicting triple-negative breast cancer ici response based on tumor mutation burden.BioMedInformatics, 5(1):9, 2025

    Houda Bendani, Nasma Boumajdi, Lahcen Belyamani, and Azeddine Ibrahimi. A decision-aid model for pre- dicting triple-negative breast cancer ici response based on tumor mutation burden.BioMedInformatics, 5(1):9, 2025

    2025

  66. [66]

    Evaluating a large language model’s ability to solve programming exercises from an introductory bioinformatics course

    Stephen R Piccolo, Paul Denny, Andrew Luxton-Reilly, Samuel H Payne, and Perry G Ridge. Evaluating a large language model’s ability to solve programming exercises from an introductory bioinformatics course. PLOS Computational Biology, 19(9):e1011511, 2023

    2023

  67. [67]

    Bixbench: a comprehensive benchmark for llm- based agents in computational biology.arXiv preprint arXiv:2503.00096, 2025

    Ludovico Mitchener, Jon M Laurent, Benjamin Ten- mann, Siddharth Narayanan, Geemi P Wellawatte, Andrew White, Lorenzo Sani, and Samuel G Ro- driques. Bixbench: a comprehensive benchmark for llm- based agents in computational biology.arXiv preprint arXiv:2503.00096, 2025

    arXiv 2025

  68. [68]

    Prediction of lupus classification criteria via generative ai medical record profiling.BioTech, 14(1):15, 2025

    Sandeep Nair, Gerald H Lushington, Mohan Pu- rushothaman, Bernard Rubin, Eldon Jupe, and San- tosh Gattam. Prediction of lupus classification criteria via generative ai medical record profiling.BioTech, 14(1):15, 2025

    2025

  69. [69]

    Funcfetch: an llm-assisted workflow enables mining thousands of enzyme–substrate inter- actions from published manuscripts.Bioinformatics, 41(1):btae756, 2025

    Nathaniel Smith, Xinyu Yuan, Chesney Melissinos, and Gaurav Moghe. Funcfetch: an llm-assisted workflow enables mining thousands of enzyme–substrate inter- actions from published manuscripts.Bioinformatics, 41(1):btae756, 2025

    2025

  70. [70]

    Gene-associated disease discovery powered by large language models.arXiv preprint arXiv:2401.09490, 2024

    Jiayu Chang, Shiyu Wang, Chen Ling, Zhaohui Qin, and Liang Zhao. Gene-associated disease discovery powered by large language models.arXiv preprint arXiv:2401.09490, 2024. 36

    Pith/arXiv arXiv 2024

  71. [71]

    Jia-Ying Chen, Jing-Fu Wang, Yue Hu, Xin-Hui Li, Yu-Rong Qian, and Chao-Lin Song. Evaluating the advancements in protein language models for encoding strategies in protein function prediction: a comprehen- sive review.Frontiers in bioengineering and biotechnol- ogy, 13:1506508, 2025

    2025

  72. [72]

    Lorenzo Pantolini, Gabriel Studer, Joana Pereira, Janani Durairaj, Gerardo Tauriello, and Torsten Schwede. Embedding-based alignment: combining pro- tein language models with dynamic programming align- ment to detect structural similarities in the twilight- zone.Bioinformatics, 40(1):btad786, 2024

    2024

  73. [73]

    Fine- tuningproteinembeddingsforfunctionalsimilarityeval- uation.Bioinformatics, 40(8):btae445, 2024

    Andrew Dickson and Mohammad RK Mofrad. Fine- tuningproteinembeddingsforfunctionalsimilarityeval- uation.Bioinformatics, 40(8):btae445, 2024

    2024

  74. [74]

    learnmsa2: deep protein multiple alignments with large language and hidden markov models.Bioinformatics, 40(Supplement_2):ii79–ii86, 2024

    Felix Becker and Mario Stanke. learnmsa2: deep protein multiple alignments with large language and hidden markov models.Bioinformatics, 40(Supplement_2):ii79–ii86, 2024

    2024

  75. [75]

    Cathe: detection of remote homologues for cath superfamilies using embeddings from protein lan- guage models.Bioinformatics, 39(1):btad029, 2023

    Vamsi Nallapareddy, Nicola Bordin, Ian Sillitoe, Michael Heinzinger, Maria Littmann, Vaishali P Wa- man, Neeladri Sen, Burkhard Rost, and Christine Orengo. Cathe: detection of remote homologues for cath superfamilies using embeddings from protein lan- guage models.Bioinformatics, 39(1):btad029, 2023

    2023

  76. [76]

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

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

    2021

  77. [77]

    Biosnap datasets: Stanford biomedical network dataset collec- tion.Note: http://snap

    Marinka Zitnik, Rok Sosic, and Jure Leskovec. Biosnap datasets: Stanford biomedical network dataset collec- tion.Note: http://snap. stanford. edu/biodata Cited by, 5(1), 2018

    2018

  78. [78]

    Accurate and transferable drug–target in- teraction prediction with druglamp.Bioinformatics, 40(12):btae693, 2024

    Zhengchao Luo, Wei Wu, Qichen Sun, and Jinzhuo Wang. Accurate and transferable drug–target in- teraction prediction with druglamp.Bioinformatics, 40(12):btae693, 2024

    2024

  79. [79]

    Joanna S Amberger, Carol A Bocchini, François Schi- ettecatte, Alan F Scott, and Ada Hamosh. Omim. org: Online mendelian inheritance in man (omim®), an online catalog of human genes and genetic disorders. Nucleic acids research, 43(D1):D789–D798, 2015

    2015

  80. [80]

    Uniprotkb/swiss-prot: the manually annotated section of the uniprot knowledgebase

    Emmanuel Boutet, Damien Lieberherr, Michael Tognolli, Michel Schneider, and Amos Bairoch. Uniprotkb/swiss-prot: the manually annotated section of the uniprot knowledgebase. InPlant bioinformatics: methods and protocols, pages 89–112. Springer, 2007

    2007

Showing first 80 references.