Pith. sign in

REVIEW 5 major objections 5 minor 41 references

TextOmics-Guided Diffusion for Hit-like Molecular Generation

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that jointly conditioning diffusion-based molecular generation on gene expression profiles and textual descriptions produces hit-like molecules that beat single-modality baselines in validity, novelty, diversity, and…

desk verdict A genuinely useful heterogeneous dataset and a plausible joint-conditioning framework, but the SOTA claim rests on text prompts that paraphrase the target structure and a self-referential hit-ratio threshold. read the letter →

arxiv 2507.09982 v1 pith:NPSMBVNC submitted 2025-07-14 cs.CL

Hang Yuan , Chen Li , Wenjun Ma , Yuncheng Jiang This is my paper · ORCID
classification cs.CL
keywords hit-likemoleculargenerationconditionaldiffusionomics-guidedtext-guidedheterogeneousdataintegrationSELFIESdrugdiscoveryzero-shottherapeutic
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

This paper introduces TextOmics, a dataset that pairs omics expression profiles with molecular textual descriptions and SELFIES structures, and ToDi, a diffusion framework that conditions generation on both omics and text. The authors aim to show that jointly integrating biological and semantic signals yields chemically valid, novel, diverse, hit-like molecules—plausible starting points for drug discovery—better than conditioning on either modality alone. If true, this would let researchers generate candidate molecules directly from a cell line's or patient's transcriptional state plus a short chemical description, speeding up early-stage target-specific drug discovery.

What carries the argument

The load-bearing mechanism is a joint conditioning pipeline built on three modules: OmicsEn, a variational autoencoder that maps gene expression profiles into a latent embedding $Z_O$; TextEn, a frozen SciBERT encoder that extracts a semantic embedding $Z_D$ from molecular text; and DiffGen, a diffusion Transformer that denoises vocabulary-aware remapped SELFIES sequences under the fused condition $Z = T(X_t + PE + TE_t, Z_D) \oplus Z_O$. SELFIES, a molecular string format whose grammar guarantees chemically valid outputs, serves as the bridge representation linking omics and text to molecular structure and is what allows the reported perfect validity.

What would settle it

Regenerate the benchmark's textual descriptions using descriptions authored independently of the target molecule, or sourced from curated chemistry texts, then rerun the ToDi versus text-only and omics-only ablations on the same test split; if the joint gains in novelty and structural similarity shrink or reverse, the claimed multimodal advantage is an artifact of text re-encoding the target structure rather than a genuine integration of complementary signals.

Watch

Extended reading notes

Core claim

The paper's central claim is that a conditional diffusion generator, jointly guided by an omics expression embedding and a molecular textual description embedding, produces hit-like molecules that are more valid, more unique, more novel, and structurally closer to known ligands than generation conditioned on omics alone or text alone. The authors support this with benchmark results on their own TextOmics dataset, with transfer experiments across ten cancer-related targets, and with a zero-shot Alzheimer's disease case study in which patient symptom narratives replace molecular descriptions.

Load-bearing premise

The load-bearing assumption is that the BioT5-generated textual descriptions provide semantic information independent of the molecule's structure, yet each description is produced from the very molecule it is paired with, so the text may simply restate the structure and thereby predetermine the observed multimodal gains.

Editorial extensions

If this is right

  • On the ChemInduced test set, the full ToDi model reaches 100% validity, 98.45% uniqueness, and 97.30% novelty, improving novelty by 5.1 points over the omics-only variant and by 8.05 points over the previous omics-guided baseline.
  • Omitting omics still leaves a text-only variant that outperforms the text-diffusion baseline TGM on the ChEBI-20 benchmark across most reported metrics, indicating the text encoder alone carries substantial structural information.
  • Transferring ToDi to ten cancer-relevant targets yields 100% validity on every target and perfect MACCS Tanimoto scores on six of ten, suggesting the joint conditioning generalizes to target-specific ligand generation.
  • In the zero-shot Alzheimer's study, molecules generated under symptom-narrative guidance achieve higher Morgan and MACCS similarity to approved drugs than the prior baseline, suggesting the framework can operate without target-specific training text.
  • The ablation results imply that omics and text contribute complementary information, since each single-modality variant outperforms prior single-modality baselines while the joint model improves further.

Reading between the lines

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

  • Editorial inference: because each textual description is generated from the target molecule itself via BioT5, the text condition may largely re-encode the molecular structure rather than supply independent biological or semantic knowledge; the reported multimodal advantage could be an artifact of this text–structure coupling.
  • Editorial inference: a fairer test of genuine multimodal synergy would pair omics profiles with human-written or externally sourced descriptions for held-out molecules, or shuffle the text–molecule pairing to see whether the joint gains persist when the text is not derived from the target structure.
  • Editorial inference: the zero-shot therapeutic claim is limited to a single case study; extending it to rare diseases would require new disease-specific omics datasets and experimental validation of the generated candidates before any therapeutic conclusion can be drawn.
  • Editorial inference: the omics encoder's reconstruction fidelity, shown by PCA overlap, establishes that the latent represents the input profile, but it does not by itself establish that the omics condition carries pharmacological relevance beyond what the text already encodes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces TextOmics, a benchmark that links omics expression profiles, molecular textual descriptions, and SELFIES molecular representations in a one-to-one manner, and proposes ToDi, a diffusion-based generative framework that conditions on both omics embeddings (OmicsEn) and text embeddings (TextEn) to generate hit-like molecules. Experiments on ChemInduced, TargetPerturb, and DiseaseSign report validity, uniqueness, novelty, Levenshtein similarity, FCD, Morgan/MACCS Tanimoto, and a hit-ratio metric, with the paper claiming state-of-the-art performance and strong zero-shot potential.

Significance. If the claims are substantiated, TextOmics and ToDi would be a useful contribution to multimodal molecular generation: the dataset is a new resource, the architecture is modular and clearly described, the code is promised to be open, and ablations (Table D.6) isolate the contribution of each modality. However, the central ``outperforms SOTA'' claim is weakened by the benchmark construction: the textual descriptions are generated from the target molecule's own SELFIES via BioT5 (Section 3.1, Appendix A.3), so the text condition is a re-encoding of the structure rather than an independent semantic source. In addition, the validity metric is predetermined by the SELFIES representation, and the hit-ratio metric in Algorithm E.1 uses a self-referential threshold. These issues do not necessarily invalidate the framework, but they must be corrected before the headline claims are credible.

major comments (5)
  1. [Section 3.1 and Appendix A.3] Every molecular textual description in TextOmics is produced by BioT5 from the SELFIES representation of the target molecule and then manually verified. Consequently, the text condition is a lossy paraphrase of the very structure the model is asked to generate. The improvement of full ToDi over ToDiw/o T in Tables 2 and D.6 can therefore be explained by the cross-attention mechanism reading functional-group tokens directly from the text, without the model learning any independent biological or semantic association. Please address this confound, for example by evaluating on independently authored descriptions (e.g., ChEBI-20 natural-language captions) or by a control experiment in which the text condition is perturbed while the target structure remains fixed, and show that the multimodal gain is not purely a structure-reconstruction artifact.
  2. [Section 4.2, Eq. (E.3), and Algorithm E.1] The hit-ratio metric defines the threshold δ as the average noise estimation error across the model's own generated samples, and counts a molecule as a hit if its noise error is below this self-derived threshold OR if a functional-group match succeeds. This makes the hit ratio self-referential: a model whose errors are tightly concentrated can inflate its hit ratio without any semantic alignment. Please define δ independently of the model's output distribution, or better, compute the hit ratio purely from functional-group matching, and report the functional-group match and noise-error criteria separately.
  3. [Table 2 and Section 4.3] The reported 100% validity scores are a property of the SELFIES representation, not a learned achievement, because every SELFIES string decodes to a valid molecular graph (as the paper itself states in Appendix A.1). Presenting this value as a distinguishing model result in Table 2 and in the surrounding discussion overstates the contribution. The validity column should either be removed from the comparison or replaced by a more informative check, such as validity after conversion to SMILES and RDKit sanitization, with a note that SELFIES guarantees syntactic validity by construction.
  4. [Section 4.5, Table 3] After transferring ToDi from ChemInduced to TargetPerturb, the model achieves a MACCS Tanimoto score of 1.00 on six of ten target proteins. Since the model was not trained on TargetPerturb molecules, perfect MACCS agreement with the reference ligands is surprising and needs explanation. Please clarify how the reference ligands for each target were selected, whether there is any overlap with the ChemInduced training set, and report how many of the generated molecules are exact or near-duplicates of the reference molecules. Without this, the cross-target generalization claim in Table 3 is not fully supported.
  5. [Section 4.6 and Section 5] The zero-shot therapeutic-generation claim rests on similarity scores for two Alzheimer's disease examples (Memantine and Donepezil), and the limitation section explicitly states that the generated molecules have not undergone experimental validation and that the evaluation is confined to a case study. The abstract's phrase "remarkable potential in zero-shot therapeutic molecular generation" is disproportionate to two case-study similarity numbers. Please either expand the zero-shot evaluation to a larger set of disease queries with statistical testing, or temper the claim in the abstract and conclusion.
minor comments (5)
  1. [Section 2] In the paragraph on MolT5, "hider" should be "hinder".
  2. [Section 3.2] The DiffGen paragraph contains "molecular stings" and should read "molecular strings".
  3. [Section 4.2] The phrase "atomic-level chemcial environments" contains a typo: "chemcial" should be "chemical".
  4. [Section 3.2] The sentence fragment "donoted as w ⊂ C" should be "denoted as w ⊂ C".
  5. [Abstract and Section 4.3] The abstract and Section 4.3 state that ToDi "outperforms existing SOTA approaches" without noting that the comparison is modality-conditional; Table D.6 shows that the text-only variant ToDiw/o O has lower novelty than GxV AEs on the ChemInduced benchmark, so the headline claim should be qualified to the multimodal setting.

Circularity Check

2 steps flagged · score 4.0 of 10

The TextOmics text condition is generated from the target molecule's own SELFIES and the hit-ratio threshold is defined by the model's own average noise error, but the core generative pipeline retains independent content via external ChEBI-20 evaluation.

  1. self definitional [Section 3.1 and Appendix A.3]
    "BioT5 [37] is then employed to convert SELFIES representations into corresponding textual descriptions. To ensure the accuracy and reliability of such descriptions, all outputs are manually verified by domain experts in chemistry."

    Each molecular textual description is produced from the SELFIES of the very molecule the model is asked to generate. Thus the text condition ZD is a lossy re-encoding of the target X0, not an independent source of semantic or biological knowledge. The reported advantage of full ToDi over ToDiw/o T (Tables 2 and D.6) is therefore partly predetermined: the model can recover target structure by reading structural tokens such as 'ether', 'acid', and 'ester' from the prompt. The claim that textual guidance provides 'complementary' information beyond omics is confounded with benchmark construction, because the one-to-one correspondence is defined by this text-from-molecule generation.

  2. self definitional [Section 4.2 and Appendix E.3 (Algorithm E.1)]
    "A generated molecule x̂i is considered a hit if it exhibits the target functional group specified in the input description. In practice, we determine this by comparing ... the associated noise estimation error ||ϵ̂θ − ϵ||2^2 against a predefined threshold δ ... δ is the average noise error across all generated samples."

    The hit threshold δ is set to the average noise-estimation error of the model's own generated samples. Whether a molecule is a 'hit' therefore depends on the distribution of the quantity being measured: a model whose errors are tightly clustered around its mean will satisfy the criterion more often by construction. This makes the hit-ratio comparisons in Figure 2(b) and Figure F.8 self-referential as evidence of semantic alignment, even though the metric is presented as an absolute measure of text-molecule correspondence.

full rationale

The core generative derivation is not circular: the OmicsEn VAE objective (Eq. 1), the text-encoding objective (Eq. 2), and the diffusion objective with reconstruction, NLL, and cosine alignment (Eq. 4) are standard supervised losses and are not constructed to force the reported table entries. The paper also provides an external benchmark: ToDiw/o O is evaluated on ChEBI-20 (Table F.8), which is independent of the TextOmics text-from-SELFIES construction, giving the text-to-molecule component non-trivial external support. Self-citations to GxVAEs, TRIOMPHE, GxRNN, and HNN2Mol are used as baselines, not as load-bearing justifications for ToDi's design choices. However, two constructional self-references prevent a clean non-circular finding. First, the TextOmics benchmark defines each text description as a BioT5-generated paraphrase of the target molecule's SELFIES, so text-conditioned generation partly reduces to reconstructing the output from a transformed version of the output; the multimodal advantage over omics-only guidance is thus partly an artifact of target leakage into the prompt. Second, the Hit Ratio metric sets its threshold to the mean noise error over the same samples being classified, making the 'hit' definition relative to the model's own error distribution. These issues are real but partial: they affect the benchmark-interpretation and the semantic-alignment evidence, not all reported structural gains, and they are explicitly bounded by the paper's own limitation statement that the molecules have not undergone experimental validation and the Alzheimer's evaluation is only a case study. Overall circularity is moderate, not total.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central result depends on the fidelity of the constructed correspondences (BioT5 text, LINCS profiles, SELFIES bridge) and on the choice of alignment weight lambda, plus the validity guarantee of SELFIES. No new physical entity is postulated.

free parameters (3)
  • lambda (alignment weight) = 0.3
    Selected by sweeping values from 0.0 to 1.0 on the ChemInduced test set (Table C.5), not on a held-out validation split. It controls the cosine alignment term in Eq. (4).
  • beta (KL weight in OmicsEn) = not reported
    Appears in Eq. (1) as a scaling factor for the KL divergence term in the omics VAE, but no value is given anywhere in the paper, so the omics latent behavior is underspecified.
  • delta (hit-ratio threshold) = average noise error of generated samples
    In Algorithm E.1 and Eq. (E.3), a generated molecule counts as a hit if its noise estimation error is below delta, where delta is defined as the average noise error across all generated samples. This threshold is derived from the model's own outputs rather than fixed independently.
assumptions (4)
  • standard math Every SELFIES string decodes to a chemically valid molecule.
    Section A.1 relies on this proven property of SELFIES (Krenn et al.) to claim 100% validity, which makes the validity metric a property of the representation rather than of the learned generator.
  • domain assumption LINCS L1000 and CREEDS expression profiles faithfully represent the biological state induced by molecules, target perturbations, or disease.
    Section A.2 builds all three benchmark subsets on these transcriptomic databases and assumes that latent features extracted from them are biologically relevant for hit-like molecular design.
  • domain assumption BioT5-generated molecular textual descriptions, after manual verification, accurately and independently capture the chemical semantics of the molecules.
    Section 3.1 and Appendix A.3 state that descriptions are generated by BioT5 and verified by domain experts, but no verification artifacts, inter-annotator agreement, or quality statistics are provided. This assumption underlies the text-conditioning evaluation.
  • domain assumption Frozen SciBERT embeddings provide chemically meaningful semantic guidance for molecular generation.
    Section 3.2 uses a frozen SciBERT encoder as TextEn and assumes its contextualized representations align with molecular structure well enough to guide diffusion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TextOmics-Guided Diffusion for Hit-like Molecular Generation." pith.science (2026). https://pith.science/paper/NPSMBVNC

@misc{pith2026250709982,
  author       = {Pith},
  title        = {Pith review of: TextOmics-Guided Diffusion for Hit-like Molecular Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NPSMBVNC}},
  note         = {Machine review of arXiv:2507.09982}
}
read the original abstract

Hit-like molecular generation with therapeutic potential is essential for target-specific drug discovery. However, the field lacks heterogeneous data and unified frameworks for integrating diverse molecular representations. To bridge this gap, we introduce TextOmics, a pioneering benchmark that establishes one-to-one correspondences between omics expressions and molecular textual descriptions. TextOmics provides a heterogeneous dataset that facilitates molecular generation through representations alignment. Built upon this foundation, we propose ToDi, a generative framework that jointly conditions on omics expressions and molecular textual descriptions to produce biologically relevant, chemically valid, hit-like molecules. ToDi leverages two encoders (OmicsEn and TextEn) to capture multi-level biological and semantic associations, and develops conditional diffusion (DiffGen) for controllable generation. Extensive experiments confirm the effectiveness of TextOmics and demonstrate ToDi outperforms existing state-of-the-art approaches, while also showcasing remarkable potential in zero-shot therapeutic molecular generation. Sources are available at: https://github.com/hala-ToDi.

Figures

Figures reproduced from arXiv: 2507.09982 by the authors.

Figure 1
Figure 1. Overview of the TextOmics heterogeneous dataset and the ToDi framework for hit-like molecular generation. (a) A molecule from TextOmics induces a specific omics expression and is paired with a corresponding molecular textual description, with SELFIES serving as a bridge between these two heterogeneous representations. (b) OmicsEn extracts features from the omics expression and projects them into a latent space to ob… view at source ↗
Figure 2
Figure 2. Functional group distribution in molecular textual descriptions and corresponding hit ratio [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Detailed illustration of ToDi’s performance in terms of biological similarity and textual hit under both overexpression and knockdown conditions. contrast, SOTA baselines, limited to single-source inputs, were evaluated under the same setting. This setup illustrates the cross-task transferability of ToDi and its robustness under multi-source scenarios [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Evaluation of ToDi’s performance in zero-shot generation of biologically meaningful hit-like molecules, guided by molecular textual descriptions or patient symptom narratives. 4.6 Therapeutic Molecular Generation: Can ToDi Perform Zero-shot for Specific Diseases? In re…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 37 canonical work pages

  1. [1]

    Molecular targeted therapy for anticancer treatment

    Hye-Young Min and Ho-Young Lee. Molecular targeted therapy for anticancer treatment. Experimental & Molecular Medicine, 54(10):1670–1694, 2022

    work page 2022

  2. [2]

    Revolutionizing cancer care strategies: immunotherapy, gene therapy, and molecular targeted therapy

    Aasma Zafar, Muhammad Jawad Khan, Junaid Abu, and Aisha Naeem. Revolutionizing cancer care strategies: immunotherapy, gene therapy, and molecular targeted therapy. Molecular Biology Reports, 51(1):219, 2024

    work page 2024

  3. [3]

    What is cancer chemotherapy? Acta Oncologica, 40(2-3):166–174, 2001

    Peter Nygren. What is cancer chemotherapy? Acta Oncologica, 40(2-3):166–174, 2001

    work page 2001

  4. [4]

    From data to cure: A comprehensive exploration of multi-omics data analysis for targeted therapies

    Arnab Mukherjee, Suzanna Abraham, Akshita Singh, S Balaji, and KS Mukunthan. From data to cure: A comprehensive exploration of multi-omics data analysis for targeted therapies. Molecular Biotechnology, 67(4):1269–1289, 2025

    work page 2025

  5. [5]

    Updated insights on clinical diagnosis and targeted therapy of acute myeloid leukaemia (AML): A molecular approach

    Krishnendu Adhikary, Krishnendu Ganguly, Nirban Roy, Parimal Bar, Sonalika Mahapatra, Rajkumar Maiti, et al. Updated insights on clinical diagnosis and targeted therapy of acute myeloid leukaemia (AML): A molecular approach. Chemical Biology Letters, 12(2):1262–1262, 2025

    work page 2025

  6. [6]

    Rational molecular editing: a new paradigm in drug discovery

    Chunhua Ma, Craig W Lindsley, Junbiao Chang, and Bin Yu. Rational molecular editing: a new paradigm in drug discovery. Journal of Medicinal Chemistry, 67(14):11459–11466, 2024

    work page 2024

  7. [7]

    Hit me with your best shot: Integrated hit discovery for the next generation of drug targets

    S Neha Ashraf, J Henry Blackwell, Geoffrey A Holdgate, Simon CC Lucas, Alisa Solovyeva, R Ian Storer, and Benjamin C Whitehurst. Hit me with your best shot: Integrated hit discovery for the next generation of drug targets. Drug Discovery Today, 29(10):104143, 2024

    work page 2024

  8. [8]

    Deep generative molecular design reshapes drug discovery

    Xiangxiang Zeng, Fei Wang, Yuan Luo, and Seung-Gu Kang. Deep generative molecular design reshapes drug discovery. Cell Reports Medicine, 3(10):100850, 2022

    work page 2022

Show all 41 references
  1. [9]

    GexMolGen: cross-modal generation of hit-like molecules via large language model encoding of gene expression signatures

    Jiabei Cheng, Xiaoyong Pan, Yi Fang, Kaiyuan Yang, Yiming Xue, Qingran Yan, and Ye Yuan. GexMolGen: cross-modal generation of hit-like molecules via large language model encoding of gene expression signatures. Briefings in Bioinformatics, 25(6):bbae525, 2024

    2024

  2. [10]

    GxV AEs: Two joint V AEs generate hit molecules from gene expression profiles

    Chen Li and Yoshihiro Yamanishi. GxV AEs: Two joint V AEs generate hit molecules from gene expression profiles. Proceedings of the AAAI Conference on Artificial Intelligence , 38(12):13455–13463, 2024

    2024

  3. [11]

    Transcriptionally conditional recurrent neural network for de novo drug design

    Yuki Matsukiyo, Atsushi Tengeiji, Chen Li, and Yoshihiro Yamanishi. Transcriptionally conditional recurrent neural network for de novo drug design. Journal of Chemical Information and Modeling, 64(15):5844–5852, 2024

    2024

  4. [12]

    Machine learning-aided generative molecular design

    Yuanqi Du, Arian R Jamasb, Jeff Guo, Tianfan Fu, Charles Harris, Yingheng Wang, Chenru Duan, Pietro Li`o, Philippe Schwaller, and Tom L Blundell. Machine learning-aided generative molecular design. Nature Machine Intelligence, 6(6):589–604, 2024

    2024

  5. [13]

    De novo generation of hit-like molecules from gene expression profiles via deep learning

    Chen Li, Yuki Matsukiyo, and Yoshihiro Yamanishi. De novo generation of hit-like molecules from gene expression profiles via deep learning. arXiv preprint arXiv:2412.19422, 2024

    2024 arXiv

  6. [14]

    Text-guided molecule generation with diffusion language model

    Haisong Gong, Qiang Liu, Shu Wu, and Liang Wang. Text-guided molecule generation with diffusion language model. Proceedings of the AAAI Conference on Artificial Intelligence , 38(1):109–117, 2024

    2024

  7. [15]

    Network-based multi-omics integrative analysis methods in drug discovery: a systematic review

    Wei Jiang, Weicai Ye, Xiaoming Tan, and Yun-Juan Bao. Network-based multi-omics integrative analysis methods in drug discovery: a systematic review. BioData Mining, 18(1):27, 2025

    2025

  8. [16]

    Advances in integrated multi-omics analysis for drug-target identification

    Peiling Du, Rui Fan, Nana Zhang, Chenyuan Wu, and Yingqian Zhang. Advances in integrated multi-omics analysis for drug-target identification. Biomolecules, 14(6):692, 2024. 10

    2024

  9. [17]

    Spatial integration of multi-omics single-cell data with simo

    Penghui Yang, Kaiyu Jin, Yue Yao, Lijun Jin, Xin Shao, Chengyu Li, Xiaoyan Lu, and Xiaohui Fan. Spatial integration of multi-omics single-cell data with simo. Nature Communications, 16(1):1265, 2025

    2025

  10. [18]

    De novo generation of hit-like molecules from gene expression signatures using artificial intelligence

    Oscar M´endez-Lucio, Benoit Baillif, Djork-Arn´e Clevert, David Rouqui´e, and Joerg Wichard. De novo generation of hit-like molecules from gene expression signatures using artificial intelligence. Nature Communications, 11(1):10, 2020

    2020

  11. [19]

    Generative adversarial nets

    Ian J Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. Advances in Neural Information Processing Systems, 2:2672–2680, 2014

    2014

  12. [20]

    SMILES, a chemical language and information system

    David Weininger. SMILES, a chemical language and information system. 1. introduction to methodology and encoding rules. Journal of Chemical Information and Computer Sciences, 28(1):31–36, 1988

    1988

  13. [21]

    TRIOMPHE: Transcriptome-based inference and generation of molecules with desired phenotypes by machine learning

    Kazuma Kaitoh and Yoshihiro Yamanishi. TRIOMPHE: Transcriptome-based inference and generation of molecules with desired phenotypes by machine learning. Journal of Chemical Information and Modeling, 61(9):4303–4320, 2021

    2021

  14. [22]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. Auto-encoding variational bayes. In International Conference on Learning Representations, 2014

    2014

  15. [23]

    Recurrent neural network regularization

    Wojciech Zaremba, Ilya Sutskever, and Oriol Vinyals. Recurrent neural network regularization. arXiv preprint arXiv:1409.2329, 2014

    2014 arXiv

  16. [24]

    Functional annotation of proteins for signaling network inference in non-model species

    Lisa Van den Broeck, Dinesh Kiran Bhosale, Kuncheng Song, C´assio Flavio Fonseca de Lima, Michael Ashley, Tingting Zhu, Shanshuo Zhu, Brigitte Van De Cotte, Pia Neyt, Anna C Ortiz, et al. Functional annotation of proteins for signaling network inference in non-model species. N...

    2023

  17. [25]

    Interpretable prediction of drug-drug interactions via text embedding in biomedical literature

    Sunwoo Jung and Sunyong Yoo. Interpretable prediction of drug-drug interactions via text embedding in biomedical literature. Computers in Biology and Medicine, 185:109496, 2025

    2025

  18. [26]

    Text2mol: Cross-modal molecule retrieval with natural language queries

    Carl Edwards, ChengXiang Zhai, and Heng Ji. Text2mol: Cross-modal molecule retrieval with natural language queries. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 595–607, 2021

    2021

  19. [27]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. The Journal of Machine Learning Research, 21(1), 2020

    2020

  20. [28]

    Translation between molecules and natural language

    Carl Edwards, Tuan Lai, Kevin Ros, Garrett Honke, Kyunghyun Cho, and Heng Ji. Translation between molecules and natural language. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, 375–413, 2022

    2022

  21. [29]

    Diffusion-LM improves controllable text generation

    Xiang Li, John Thickstun, Ishaan Gulrajani, Percy S Liang, and Tatsunori B Hashimoto. Diffusion-LM improves controllable text generation. Advances in Neural Information Process- ing Systems, 35:4328–4343, 2022

    2022

  22. [30]

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

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

    2020

  23. [31]

    SELFIES and the future of molecular string representations

    Mario Krenn, Qianxiang Ai, Senja Barthel, Nessa Carson, Angelo Frei, Nathan C Frey, Pascal Friederich, Th´eophile Gaudin, Alberto Alexander Gayle, Kevin Maik Jablonka, et al. SELFIES and the future of molecular string representations. Patterns, 3(10), 2022

    2022

  24. [32]

    Integration of multi-omics data accelerates molecular analysis of common wheat traits

    Ning Zhang, Li Tang, Songgang Li, Lu Liu, Mengjuan Gao, Sisheng Wang, Daiying Chen, Yichao Zhao, Ruiqing Zheng, Armin Soleymaniniya, et al. Integration of multi-omics data accelerates molecular analysis of common wheat traits. Nature Communications, 16(1):2200, 2025. 11

    2025

  25. [33]

    Target-aware molecular graph generation

    Cheng Tan, Zhangyang Gao, and Stan Z Li. Target-aware molecular graph generation. Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 410–427, 2023

    2023

  26. [34]

    MolGene-E: Inverse molecular design to modulate single cell transcriptomics

    Rahul Ohlan, Raswanth Murugan, Li Xie, Mohammedsadeq Mottaqi, Shuo Zhang, and Lei Xie. MolGene-E: Inverse molecular design to modulate single cell transcriptomics. bioRxiv, 2025

    2025

  27. [35]

    Learning deep representations by mutual information estimation and maximization

    R Devon Hjelm, Alex Fedorov, Samuel Lavoie-Marchildon, Karan Grewal, Phil Bachman, Adam Trischler, and Yoshua Bengio. Learning deep representations by mutual information estimation and maximization. In International Conference on Learning Representations, 2019

    2019

  28. [36]

    SciBERT: A pretrained language model for scientific text

    Iz Beltagy, Kyle Lo, and Arman Cohan. SciBERT: A pretrained language model for scientific text. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, 3615–3620, 2019

    2019

  29. [37]

    BioT5: Enriching cross-modal integration in biology with chemical knowledge and natural language associations

    Qizhi Pei, Wei Zhang, Jinhua Zhu, Kehan Wu, Kaiyuan Gao, Lijun Wu, Yingce Xia, and Rui Yan. BioT5: Enriching cross-modal integration in biology with chemical knowledge and natural language associations. Proceedings of the 2023 Conference on Empirical Methods in Natural Languag...

    2023

  30. [38]

    Kullback-leibler divergence

    James M Joyce. Kullback-leibler divergence. International encyclopedia of statistical science, pages 720–722, 2011

    2011

  31. [39]

    Lincs Canvas Browser: interactive web app to query, browse and interrogate LINCS L1000 gene expression signatures

    Qiaonan Duan, Corey Flynn, Mario Niepel, Marc Hafner, Jeremy L Muhlich, Nicolas F Fernandez, Andrew D Rouillard, Christopher M Tan, Edward Y Chen, Todd R Golub, et al. Lincs Canvas Browser: interactive web app to query, browse and interrogate LINCS L1000 gene expression signat...

    2014

  32. [40]

    Extraction and analysis of signatures from the gene expression omnibus by the crowd

    Zichen Wang, Caroline D Monteiro, Kathleen M Jagodnik, Nicolas F Fernandez, Gregory W Gundersen, Andrew D Rouillard, Sherry L Jenkins, Axel S Feldmann, Kevin S Hu, Michael G McDermott, et al. Extraction and analysis of signatures from the gene expression omnibus by the crowd. ...

    2016

  33. [41]

    car- bamate ester

    David Rogers and Mathew Hahn. Extended-connectivity fingerprints. Journal of Chemical Information and Modeling, 50(5):742–754, 2010. 12 Appendix A Dataset Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...

    2010

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.