Pith. sign in

REVIEW 4 major objections 6 minor 40 references

Gene-R1 is a three-stage data-augmented fine-tuning framework that lets lightweight open-source LLMs match commercial reasoning models on gene set functional annotation.

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

2026-08-04 18:44 UTC pith:CIBI7CKV

load-bearing objection Gene-R1 is a genuinely useful three-stage fine-tuning pipeline for lightweight LLMs on gene set analysis, and the 8B model's ROUGE gains are real; but the semantic-similarity headline is muddied by using MedCPT both as a training reward and inside the evaluation metric. the 4 major comments →

arxiv 2509.10575 v1 pith:CIBI7CKV submitted 2025-09-11 q-bio.GN cs.AI

Gene-R1: Reasoning with Data-Augmented Lightweight LLMs for Gene Set Analysis

Zhizheng Wang , Yifan Yang , Qiao Jin , Zhiyong Lu This is my paper
classification q-bio.GN cs.AI
keywords gene set analysisfunctional annotationlarge language modelsreasoningdata augmentationreinforcement learningGRPOGene Ontology
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.

The paper aims to show that a lightweight, openly available LLM can perform gene set analysis—naming the biological function of a gene list and explaining it—as well as commercial reasoning models like GPT-4 and o1. It introduces Gene-R1, a three-stage fine-tuning recipe: warm up the model on gene-centric relational knowledge, activate step-by-step reasoning on traces distilled from GPT-o1, then align the output to gold-standard labels with reinforcement learning. On 1,508 in-distribution GO gene sets, Gene-R1 with an 8-billion-parameter Llama backbone outperforms all baselines on ROUGE and semantic similarity; on 106 out-of-distribution sets it is statistically indistinguishable from GPT-4 and Llama3.3-70B. The practical significance is that institutions with sensitive or proprietary gene-expression data could deploy local, low-cost models without sacrificing annotation quality.

Core claim

Gene-R1's central claim is that a carefully ordered data-augmentation pipeline—domain-knowledge pre-training, supervised reasoning distillation filtered by similarity to gold labels, and GRPO with a soft-match reward combining MedCPT semantic similarity with longest-common-subsequence lexical overlap—can close the performance gap between open-weight models and proprietary reasoning LLMs on gene set functional annotation. The strongest in-distribution numbers occur on GO:BP, where Gene-R1 (8B) reaches ROUGE-L 0.277 and average semantic similarity 0.647 versus GPT-4's 0.184 and 0.614, with improvements reported as significant at p<0.05. On 106 out-of-distribution gene sets from NeST and MSigDB

What carries the argument

The load-bearing mechanism is the three-stage fine-tuning sequence: (1) knowledge warm-up, where the backbone is pre-trained on gene-centric relational sentences from databases so gene symbols become meaningful rather than arbitrary strings; (2) reasoning activation, where the warmed-up model is fine-tuned on GPT-o1-generated step-by-step reasoning traces, retaining only those with similarity greater than 0.7 to the gold label; and (3) task alignment, where GRPO reinforcement learning optimizes a dual reward—soft-match semantic/lexical similarity to the gold annotation and exact-match structural formatting. The soft-match reward integrates MedCPT for semantic similarity and longest common su

Load-bearing premise

The load-bearing premise is that optimizing MedCPT/LSC similarity during task-alignment training (Section 3) teaches biological reasoning rather than overfitting the same semantic encoders used in evaluation (Section 4.2); if that premise gives way, the reported semantic gap-closing is inflated, although the ROUGE gains remain partly independent support.

What would settle it

Re-run the in-distribution and out-of-distribution evaluations using a semantic similarity encoder that was never used in training (for example, a freshly fine-tuned bioencoder or a blind LLM judge), and compare Gene-R1 with GPT-4 on both ROUGE and semantic similarity. If the similarity advantage disappears while the ROUGE gains persist, the claim that task-specific reasoning closes the semantic gap is not supported.

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

If this is right

  • Gene-R1 (8B) raises ROUGE-L to 0.277 on GO:BP versus GPT-4's 0.184 and average semantic similarity to 0.647 versus 0.614, with improvements significant at p<0.05.
  • On 106 out-of-distribution gene sets, Gene-R1's similarity scores show no significant difference from GPT-4 and Llama3.3-70B, supporting generalization across gene sources.
  • Ablations show that each module contributes, and that the task-alignment stage with soft-match GRPO rewards yields the largest gains, outperforming online DPO and exact-match-only GRPO.
  • Performance gains hold across 1B, 3B, and 8B Llama backbones, with over 15% similarity-score improvement over base models on every in-distribution dataset.
  • Because the final model is small and open, it can be deployed locally, reducing the cost and privacy concerns associated with sending gene-expression data to commercial APIs.

Where Pith is reading between the lines

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

  • An implication the authors leave implicit is that the evaluation's semantic similarity may partially overlap with the training reward: the soft-match reward in Section 3 uses MedCPT and LSC, while the evaluation in Section 4.2 averages MedCPT, SentenceBERT, and SapBERT; an independent encoder would isolate how much of the reported semantic gain is genuine reasoning rather than reward overfitting.
  • A natural extension the paper does not test is shifting the same three-stage pipeline to other label spaces such as disease ontology or phenotype ontology; the authors themselves list these as open questions.
  • The cascade knowledge-injection order (knowledge warm-up followed by reasoning activation) yields only modest, mostly non-significant gains over a merged single-stage injection, suggesting that the explicit warm-up phase may be less critical than the reinforcement-learning alignment phase.
  • Because the reasoning traces are distilled from GPT-o1 and filtered by similarity, the ceiling of Gene-R1 is partly bounded by the teacher's reasoning quality and coverage; replacing those traces with independently validated reasoning annotations could raise performance further.

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. The paper introduces Gene-R1, a three-stage fine-tuning pipeline that adapts lightweight open-source LLMs (Llama 3.2 1B/3B and Llama 3.1 8B) to gene set analysis (GSA). Stage 1 (knowledge warm-up) continues pre-training on gene-centric relational sentences from public biological databases. Stage 2 (reasoning activation) supervised fine-tunes the model on GPT-o1-generated step-by-step reasoning chains for GO gene sets, retaining only examples whose proposed function is >0.7 similar to the gold label. Stage 3 (task alignment) applies GRPO with a soft-match reward combining MedCPT semantic similarity and longest-common-subsequence lexical overlap, together with an exact format match. The authors evaluate on 1,508 in-distribution GO gene sets (GO:BP, GO:MF, GO:CC) and 106 out-of-distribution gene sets (NeST, MsigDB), reporting ROUGE and semantic similarity averaged over MedCPT, SentenceBERT, and SapBERT. The headline claim is that Gene-R1(8B) substantially improves over base Llama models and matches commercial LLMs such as GPT-4 in-distribution and on out-of-distribution gene sets.

Significance. If the reported gains are robust, the paper makes a useful practical contribution: it provides a concrete recipe for distilling reasoning capability into small open-source models for a biomedical annotation task, shows a clear benefit of domain-knowledge warm-up, and demonstrates a potential path to local, privacy-preserving deployment. The breadth of evaluation across three model sizes and five datasets is a strength, and the inclusion of an out-of-distribution test is valuable. However, the central quantitative claim is currently weakened by the overlap between the training reward and the evaluation encoders, and by the absence of code, data, and seed-level variance. The underlying idea is promising, but the evidence as presented needs substantial strengthening before the 'matching commercial LLMs' conclusion can be accepted.

major comments (4)
  1. [Section 3; Section 4.2; Tables 4-5] The semantic evaluation is not independent of the training reward. In the RA module, GPT-o1 outputs are retained only when a similarity score to the gold label exceeds 0.7; in the TA module, the GRPO reward is a soft match combining MedCPT and LSC. Section 4.2 then defines the evaluation similarity as the average of MedCPT, SentenceBERT, and SapBERT, and ROUGE-L is itself based on longest common subsequence. Thus Gene-R1 is directly optimized against the same family of similarity functions used in evaluation. The reported 'Score (avg.)' in Table 4 and the 'n.s.' comparisons in Table 5 may be driven primarily by MedCPT, with SentenceBERT and SapBERT contributing little or nothing. The authors should report per-encoder scores for MedCPT, SentenceBERT, and SapBERT separately, run significance tests per encoder, and ideally add a held-out evaluation encoder that was not used in RA/TA. Withou
  2. [Section 5.1; Table 4] The text overstates the consistency of the results. Table 4 shows that Gene-R1(1B) on GO:BP has a semantic similarity of 0.617, below GPT-4o's 0.629, and Gene-R1(3B) on GO:CC is 0.578, below several baselines (e.g., o1 at 0.609). Moreover, Table 6 shows Gene-R1(3B) produces correct output format in only 42.4% of GO:BP and 42.9% of GO:CC cases. The abstract and Section 5.1 claim that Gene-R1 'consistently outperforms all baselines' and 'closes the performance gap' with commercial LLMs; this is defensible only for the 8B variant in-distribution. Please qualify the claims by model size, report per-variant significance tests, and directly address the 3B format-instability issue.
  3. [Section 4.2; Section 4.3; Tables 4-5] No code, training data, or evaluation scripts are provided, and all results are reported as single numbers without error bars or seed-level variance. The paired t-test p-values in Tables 4 and 5 are computed over items within one run, not over independent training seeds, so they do not establish stability of the pipeline. Since the paper's contribution is a training recipe, this is a major reproducibility gap. The authors should release the RA/TA datasets, the filtering and reward code, and evaluation scripts, and report mean and standard deviation over at least three seeds for the main comparisons.
  4. [Section 3; Section 5.3] The similarity threshold of 0.7 used to filter GPT-o1-generated reasoning examples is not defined: which encoder computes this similarity? This is important both because it determines the quality of the SFT corpus and because it is directly related to the evaluator-overlap issue. In addition, no sensitivity analysis is provided for this threshold, nor for the relative weights of MedCPT and LSC in the soft-match reward. At minimum, specify the encoder and report how performance varies as the threshold and reward weights are changed.
minor comments (6)
  1. [General] There are numerous typos and formatting errors: 'realtional', 'modue', 'acorss', 'solation', 'constract', 'bechmark', 'Alation experiments', 'incoporating', 'diffuclt', and 'specificized'. The paper needs a careful proofreading pass.
  2. [Section 4.2] The evaluation section states that similarity is averaged over three encoders to 'mitigate potential bias inherent to a single semantic encoder.' That rationale is undermined by the fact that one of the three encoders (MedCPT) is used directly in the training reward. Please move MedCPT out of the primary evaluation average or treat it as a development metric.
  3. [Section 5.1; Table 4] The 3B variant shows a striking performance drop on GO:CC (ROUGE-L 0.143 vs 0.222 for the 1B variant), which is not explained in the text. If this is due to format instability (Table 6), the reader should be told explicitly.
  4. [References; Section 4.3] The backbone is Llama 3.1 and Llama 3.2, but the paper cites Touvron et al. (2023), which introduces Llama 1. Please cite the appropriate Llama 3 model card (e.g., Grattafiori et al., 2024) or the relevant model release.
  5. [Section 6; Limitations] The limitations paragraph is honest but brief. It acknowledges manual data collection and hallucination risk on unseen genes, but it does not mention the evaluator-reward overlap or the absence of public code/data. Please add these as explicit limitations.
  6. [Figure 3] The text references results in Figure 3 without providing axis labels or error bars in the figure description. Please ensure all subfigures are labeled and that the figure caption states what is plotted and with what error bars.

Circularity Check

1 steps flagged

Semantic similarity evaluation is not independent of the training reward: MedCPT is optimized during training and included in the evaluation average.

specific steps
  1. fitted input called prediction [Section 3 (Task Alignment module, soft-match reward) and Section 4.2 (Evaluation Metrics)]
    "we implement a soft-match reward function that captures both semantic similarity and sequence overlap between f_gi and a_i, which integrates the MedCPT 36 score for semantic similarity and the longest common subsequence (LSC) score37 for lexical alignment. ... we calculated semantic similarity using three biomedical encoders — MedCPT, SentenceBERT, and SapBERT—and reported the average scores."

    The GRPO soft-match reward directly optimizes MedCPT semantic similarity to the gold label, and the evaluation semantic similarity is the average of MedCPT, SentenceBERT, and SapBERT. MedCPT is therefore both the training objective and one component of the reported metric, making that component a fitted prediction rather than a held-out measure. Because the paper reports only the average, the semantic-similarity improvement of Gene-R1 over baselines may be driven largely by the optimized MedCPT encoder. This is a partial train/eval overlap: the other two encoders and ROUGE provide independent evidence, but the specific claim of 'matching commercial LLMs' on semantic similarity is not cleanly supported without per-encoder results.

full rationale

This paper is an empirical pipeline, so circularity appears as train/eval overlap rather than equations. The central issue is that the TA module's GRPO soft-match reward uses MedCPT similarity against gold labels, and the evaluation reports the average of MedCPT, SentenceBERT, and SapBERT semantic similarity. The model is thus explicitly optimized to maximize one component of the evaluator. The RA module's similarity filter (>0.7) likely compounds this, though the paper does not specify the encoder used there. The ROUGE gains, however, are independent and substantial, and the two non-MedCPT encoders are not in the training reward, so the circularity is partial. The paper also reports no per-encoder breakdowns, making the semantic claim unverifiable. There are no load-bearing self-citations: references to the authors' own prior work (GeneAgent, GeneGPT, Cell-o1, safety paper) are contextual or dataset-provenance only. Thus the score is 6, reflecting one significant train/eval overlap that weakens a part of the central claim, but the overall pipeline is not entirely circular.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The training pipeline relies on database labels as ground truth, on three semantic encoders as valid proxies for biological correctness, on GPT-o1-generated chains as high-quality reasoning supervision, and on the cascade of domain pre-training before SFT/RL. The first two are domain assumptions; the last two are specific to this paper and not independently validated.

free parameters (3)
  • RA data filtering similarity threshold = 0.7
    Only GPT-o1 outputs with similarity > 0.7 to the gold label are kept for supervised reasoning training; the value is chosen by hand and never ablated.
  • Soft-match reward weights (MedCPT vs LSC) = unspecified
    TA module combines MedCPT semantic similarity and LSC lexical overlap into one reward, but the combination weights and the relative weight of the exact-match format reward are not reported.
  • Training hyperparameters (learning rate, epochs, GRPO steps) = unspecified
    The central result depends on these choices, but they are not given in the implementation section.
axioms (4)
  • domain assumption GO and database gold-standard labels are correct functional annotations for the evaluation gene sets.
    The whole evaluation treats labels from GO, NeST, and MsigDB as ground truth; annotation errors would shift all scores.
  • domain assumption The semantic encoders MedCPT, SentenceBERT, and SapBERT measure biological plausibility of generated function names.
    Section 4.2 defines the main non-lexical metric as the average of these encoders; if embeddings reward paraphrase rather than biological correctness, scores can be high for wrong functions.
  • ad hoc to paper GPT-o1-generated reasoning chains, after >0.7 similarity filtering, are high-quality supervision for biological reasoning.
    Section 3 RA module relies on this without manual validation; noisy reasoning would be learned by the student.
  • ad hoc to paper Gene-centric relational pre-training sentences are sufficient to teach gene-symbol grounding and do not introduce conflicting representations.
    Section 3 KW hypothesis is tested only indirectly through end-task gains.

pith-pipeline@v1.3.0-alltime-deepseek · 12071 in / 15796 out tokens · 165529 ms · 2026-08-04T18:44:54.055772+00:00 · methodology

0 comments
read the original abstract

The gene set analysis (GSA) is a foundational approach for uncovering the molecular functions associated with a group of genes. Recently, LLM-powered methods have emerged to annotate gene sets with biological functions together with coherent explanatory insights. However, existing studies primarily focus on proprietary models, which have been shown to outperform their open-source counterparts despite concerns over cost and data privacy. Furthermore, no research has investigated the application of advanced reasoning strategies to the GSA task. To address this gap, we introduce Gene-R1, a data-augmented learning framework that equips lightweight and open-source LLMs with step-by-step reasoning capabilities tailored to GSA. Experiments on 1,508 in-distribution gene sets demonstrate that Gene-R1 achieves substantial performance gains, matching commercial LLMs. On 106 out-of-distribution gene sets, Gene-R1 performs comparably to both commercial and large-scale LLMs, exhibiting robust generalizability across diverse gene sources.

Figures

Figures reproduced from arXiv: 2509.10575 by Qiao Jin, Yifan Yang, Zhiyong Lu, Zhizheng Wang.

Figure 1
Figure 1. Figure 1: Example illustration of functional enrichment analysis (a) and LLM-based solutions (b) for the GSA task. To address this shortcoming, we propose Gene-R1, a data-augmented fine-tuning framework that endows lightweight, open-source LLMs with step-by-step reasoning capabilities tailored to the GSA task, aiming to close the performance gap with commercial reasoning LLMs. As illustrated in Fig.2, Gene-R1 compri… view at source ↗
Figure 2
Figure 2. Figure 2: The overall framework of Gene-R1, which consists of three modules designed to fine-tune lightweight Llama models for the GSA task. These modules are knowledge warm-up, reasoning activation, and task alignment. These modules play a systematic role for incorporating curated knowledge, enhancing reasoning capabilities, and refining the output format of the fine-tuned model, respectively. 2. Related Works Our … view at source ↗
Figure 3
Figure 3. Figure 3: Fig.3. As summarized in [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

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

40 extracted references · 10 linked inside Pith

  1. [1]

    de Leeuw, C., Neale, B., Heskes, T. et al. The statistical properties of gene-set analysis. Nat Rev Genet 17, 353–364 (2016)

    2016

  2. [2]

    Hu, M., Alkhairy, S., Lee, I. et al. Evaluation of large language models for discovery of gene set function. Nat Methods 22, 82–91 (2025)

    2025

  3. [3]

    Ma, K., Huang, S., Ng, K.K. et al . Saturation mutagenesis -re-inforced functional assays for disease-related genes. Cell 187, 6707–6724.e22 (2024)

    2024

  4. [4]

    Tachmazidou, I., Hatzikotoulas, K., Southam, L. et al. Identification of new therapeutic targets for osteoarthritis through genome -wide analyses of UK Biobank data. Nat Genet 51, 230–236 (2019)

    2019

  5. [5]

    Subramanian, A., Tamayo, P., Mootha, V.K. et al. Gene set enrichment analysis: a knowledge- based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences 102, 15545-15550 (2005)

    2005

  6. [6]

    Raudvere, U., Kolberg, L., Kuzmin, I. et al. g: Profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update). Nucleic acids research 47, W191-W198 (2019)

    2019

  7. [7]

    Ashburner, M., Ball, C., Blake, J. et al. Gene Ontology: tool for the unification of biology. Nat Genet 25, 25–29 (2000)

    2000

  8. [8]

    Liberzon, A., Birger, C., Thorvaldsdóttir, H. et al. The molecular signatures database hallmark gene set collection. Cell systems 1, 417-425 (2015)

    2015

  9. [9]

    Joachimiak, M.P., Caufield, J.H., Harris, N.L. et al . Gene set summarization using large language models. ArXiv, arXiv-2305 (2024)

    2024

  10. [10]

    Wang, Z., Jin, Q., Wei, CH. et al. GeneAgent: self-verification language agent for gene -set analysis using domain databases. Nat Methods 1-9 (2025)

    2025

  11. [11]

    Liu, A., Feng, B., Xue, B. et al. Deepseek-v3 technical report. arXiv:2412.19437 (2024)

    Pith/arXiv arXiv 2024

  12. [12]

    Jaech, A., Kalai, A., Lerer, A. et al. Openai o1 system card. arXiv:2412.16720 (2024)

    Pith/arXiv arXiv 2024

  13. [13]

    Yang, Y., Jin, Q., Leaman, R. et al . Ensuring safety and trust: Analyzing the risks of large language models in medicine. arXiv preprint arXiv:2411.14487 (2024)

    Pith/arXiv arXiv 2024

  14. [14]

    Wang, Z., Du, Y., Sun, Z. et al . Re2llm: reflective reinforcement large language model for session-based recommendation. Proceedings of the AAAI Conference on Artificial Intelligence 39, 12827-12835 (2025)

    2025

  15. [15]

    Trung, L., Zhang, X., Jie, Z. et al. Reft: Reasoning with reinforced fine-tuning. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics 1, 7601-7614 (2024)

    2024

  16. [16]

    Qi, B., Li, P., Li, F. et al . Online dpo: Online direct preference optimization with fast -slow chasing. arXiv preprint arXiv:2406.05534 (2024)

    Pith/arXiv arXiv 2024

  17. [17]

    Shao, Z., Wang, P., Zhu, Q. et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300 (2024)

    Pith/arXiv arXiv 2024

  18. [18]

    Xue, Z., Li, L., Tian, S. et al. Domain knowledge is all you need: A field deployment of llm - powered test case generation in fintech domain. Proceedings of the 2024 IEEE/ACM 46th International Conference on Software Engineering: Companion Proceedings 314-315 (2024)

    2024

  19. [19]

    Song, Z., Yan, B., Liu, Y. et al . Injecting domain -specific knowledge into large language models: a comprehensive survey. arXiv preprint arXiv:2502.10708 (2025)

    Pith/arXiv arXiv 2025

  20. [20]

    V., Jones, M

    Kuleshov, M. V., Jones, M. R., Rouillard, A. D. et al . Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic acids research 44, W90-W97 (2016)

    2016

  21. [21]

    Genegpt: Augmenting large language models with domain tools for improved access to biomedical information

    Jin, Q., Yang, Y., Chen, Q., & Lu, Z. Genegpt: Augmenting large language models with domain tools for improved access to biomedical information. Bioinformatics 40, btae075 (2024)

    2024

  22. [22]

    Wu, Q., Bansal, G., Zhang, J. et al. Autogen: Enabling next -gen LLM applications via multi - agent conversations. First Conference on Language Modeling (2024)

    2024

  23. [23]

    Han, T., Adams, L.C., Papaioannou, J.M. et al . MedAlpaca--an open -source collection of medical conversational AI models and training data. arXiv preprint arXiv:2304.08247 (2023)

    Pith/arXiv arXiv 2023

  24. [24]

    Toma, A., Lawler, P .R., Ba, J. et al. Clinical camel: An open expert-level medical language model with dialogue-based knowledge encoding. arXiv preprint arXiv:2305.12031 (2023)

    Pith/arXiv arXiv 2023

  25. [25]

    Singhal, K., Tu, T., Gottweis, J. et al. Toward expert-level medical question answering with large language models. Nat Med 31, 943–950 (2025)

    2025

  26. [26]

    Zhang, K., Zeng, S., Hua, E. et al. Ultramedical: Building specialized generalists in biomedicine. Advances in Neural Information Processing Systems 37, 26045-26081 (2024)

    2024

  27. [27]

    Fang, Y., Jin, Q., Xiong, G. et al . Cell-o1: Training LLMs to Solve Single -Cell Reasoning Puzzles with Reinforcement Learning. arXiv preprint arXiv:2506.02911 (2025)

    Pith/arXiv arXiv 2025

  28. [28]

    Davis, A.P., Grondin, C.J., Johnson, R.J. et al. Comparative toxicogenomics database (CTD): update 2021. Nucleic acids research 49, D1138-D1143 (2021)

    2021

  29. [29]

    Annotation of biologically relevant ligands in UniProtKB using ChEBI

    Coudert, E., Gehant, S., De Castro et al . Annotation of biologically relevant ligands in UniProtKB using ChEBI. Bioinformatics 39, btac793 (2023)

    2023

  30. [30]

    Fabregat, A., Jupe, S., Matthews, L. et al. The reactome pathway knowledgebase. Nucleic acids research 46, D649-D655 (2018)

    2018

  31. [31]

    Agrawal, A., Balcı, H., Hanspers, K. et al . WikiPathways 2024: next generation pathway database. Nucleic acids research 52, D679-D689 (2024)

    2024

  32. [32]

    Thomas, P.D., Ebert, D., Muruganujan, A. et al . PANTHER: Making genome ‐ scale phylogenetics accessible to all. Protein Science 31, 8-22 (2022)

    2022

  33. [33]

    Giurgiu, M., Reinhard, J., Brauner, B. et al . CORUM: the comprehensive resource of mammalian protein complexes—2019. Nucleic acids research 47, D559-D563 (2019)

    2019

  34. [34]

    Geer, L.Y., Marchler-Bauer, A., Geer, R.C. et al. The NCBI biosystems database. Nucleic acids research 38, D492-D496 (2010)

    2010

  35. [35]

    Zheng, F., Kelly, M.R., Ramms, D.J. et al. Interpretation of cancer mutations using a multiscale map of protein systems. Science 374, eabf3067 (2021)

    2021

  36. [36]

    Jin, Q., Kim, W., Chen, Q. et al. Medcpt: Contrastive pre-trained transformers with large-scale pubmed search logs for zero-shot biomedical information retrieval. Bioinformatics 39, btad651 (2023)

    2023

  37. [37]

    Rouge: A package for automatic evaluation of summaries

    Lin, C.Y. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, 74-81 (2004)

    2004

  38. [38]

    Sentence -BERT: Sentence Embeddings using Siamese BERT - Networks

    Reimers, N., & Gurevych, I. Sentence -BERT: Sentence Embeddings using Siamese BERT - Networks. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 3982-3992 (2019)

    2019

  39. [39]

    Liu, F., Shareghi, E., Meng, Z. et al . Self-alignment pretraining for biomedical entity representations. North American Association for Computational Linguistics 2021 , Association for Computational Linguistics (ACL), 4228-4238 (2021)

    2021

  40. [40]

    Touvron, H., Lavril, T., Izacard, G. et al . Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023)

    Pith/arXiv arXiv 2023