PKAG-DDI: Pairwise Knowledge-Augmented Language Model for Drug-Drug Interaction Event Text Generation

Feng Huang; Wen Zhang; Zhankun Xiong; Ziyan Wang

REVIEW 2 major objections 5 minor 40 references

PKAG-DDI: Pairwise Knowledge-Augmented Language Model for Drug-Drug Interaction Event Text Generation

T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Pairwise drug roles improve AI-written drug-interaction texts.

desk verdict Worth reading and worth refereeing, but the 'especially in inductive scenarios' claim is thinner than the abstract suggests once you look at the MecDDI rows. read the letter →

arxiv 2507.19011 v1 pith:FIN4M2WE submitted 2025-07-25 q-bio.BM

Ziyan Wang , Zhankun Xiong , Feng Huang , Wen Zhang This is my paper · ORCID

classification q-bio.BM

keywords drug-druginteractioneventtextgenerationbiologicalfunctionpairwiseknowledgeselectionretrieval-augmentedlanguagemodelmoleculargraphinductivescenario

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that drug-drug interaction event (DDIE) text is generated more accurately when the language model knows, for each of the two drugs, which biological function of that drug is actually engaged in the interaction. To supply that knowledge without manual annotation, the proposed model PKAG-DDI first predicts the most relevant biological function of each drug conditioned on the other drug, then feeds the selected function pairs into the generator by marginalizing over all plausible pairings. On two professional databases, the generated texts score higher on BLEU, ROUGE, and METEOR than generation-based baselines, and the advantage grows in inductive settings where test drugs were unseen during training. An upper-bound version that receives the gold biological functions achieves near-perfect generation scores, indicating that the selected functions carry most of the useful signal.

What carries the argument

The load-bearing mechanism is the pairwise knowledge selector (PKS) together with the pairwise knowledge integration strategy. PKS builds drug representations from fingerprints and molecular-graph node prototypes, uses cross-attention to let each drug condition the other's representation in a single shared computation, and predicts each drug's top-K biological functions. The integration strategy constructs all K×K pairs, assigns each pair the product of the two renormalized top-K probabilities, and generates the DDIE text token by token by marginalizing over these pairs, turning knowledge selection into a probabilistic part of generation rather than a rigid prompt addition.

What would settle it

Retrain the pairwise knowledge selector with randomly chosen biological functions as gold labels, keeping the generator fixed; if PKAG-DDI still beats the baselines by the same margin, biological-function selection is not the source of the improvements. Alternatively, feed the generator deliberately mismatched function pairs and test whether the generated texts degrade in a blinded human evaluation.

Watch

Extended reading notes

Core claim

The central discovery is that biological function is a pairwise, interaction-specific property: the relevant role of one drug can only be decided in the presence of the other. PKAG-DDI represents both drugs as molecular graphs and injects each drug's node prototypes into the other via cross-attention, producing representations from which a classifier selects the top-K biological functions for each direction. The generator does not simply concatenate these functions into the prompt; it treats them as latent variables and computes each token's probability as a weighted sum over all K×K function pairs, with weights given by the joint selector distribution. The paper argues this avoids the noise of dumping every function into the prompt and prevents the mismatches that arise from ranking-based pairing. The full model outperforms two generation baselines on both the MecDDI and DDInter2.0 datasets under random, cold-start, and scaffold splits, and it does so without sacrificing classification accuracy relative to dedicated classifiers.

Load-bearing premise

The load-bearing premise is that the BM25-selected gold biological functions used to train the selector are correct enough; if these labels are noisy, selection errors propagate into the generated interaction text.

Editorial extensions

If this is right

DDIE prediction can be formulated as open-ended text generation instead of label classification, giving clinicians detailed mechanism descriptions without needing a label-to-text dictionary.
The near-perfect scores of the gold-function upper bound imply that once the correct pairwise biological functions are known, the remaining text-generation task is almost solved on the datasets studied.
Because the selected biological functions are visible before the text is written, the model offers an interpretable intermediate step showing why an interaction is expected.
The probabilistic marginalization over knowledge pairs is a general recipe for relational text generation where two entities jointly determine an outcome.

Reading between the lines

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

If the biological-function selector were raised to near-oracle accuracy, generation scores would likely approach the gold-function upper bound; the gap between the full model and that bound measures how much error comes from knowledge selection rather than language generation.
The use of BM25 to pick gold labels ties the selector to lexical overlap between function names and event text; a semantically or causally grounded label choice might improve the inductive scenarios, where the selector currently weakens most.
A testable extension would swap the MecDDI knowledge source for another structured drug-knowledge base while keeping the same architecture; persistent gains would show the method exploits pairwise context in general, not the specific vocabulary of one database.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Here's my read. The genuinely new bit is using the drug pair's biological functions as conditional context for DDIE text generation, and the two-stage design—cross-attention selector with weight reuse, then marginalization over top-K biological-function pairs—is a coherent combination I haven't seen in the cited work. The experiments are broad: two datasets, three splits, generation and classification metrics, ablations, and an oracle upper bound. The oracle experiment is the most useful result: with gold biological functions, PKAG-DDI* hits ~99% on MecDDI, which strongly supports the premise that biological functions carry the information the generator needs. The released code and data make this reproducible.

The soft spot the stress-test note identifies is real. The abstract says the method wins 'especially in challenging inductive scenarios,' but on MecDDI the inductive margins are tiny or negative: PKAG-DDI trails MolT5 on METEOR and ROUGE-L in the scaffold split, and the cold-start lead over MolT5 is 0.05 METEOR points. The clear inductive wins are on DDInter2.0 (e.g., +6.8 BLEU-2 cold start), so the claim should be dataset-dependent. Since all numbers are means of three runs with no standard deviations or significance tests, the reader genuinely cannot tell whether the MecDDI differences are noise. That is a reporting gap, not a fraudulent result. I would not call the inductive claim load-bearing enough to reject; but it needs to be hedged or backed by variance estimates.

The BM25-label issue is real but secondary. Training the selector on labels chosen by lexical similarity to the target DDIE text does create a mild circular signal, and the authors admit in Appendix B that they chose single-label classification to reduce error propagation. At inference the target text isn't used, so the generation comparison is not circular. Still, PKS top-2 accuracy is 25.8% on the MecDDI scaffold split, which makes the small inductive gains even harder to interpret.

Minor: leaving ZeroDDI out of the classification baselines is odd, since they cite it; including it would strengthen the comparison. The 'first to use biological functions' claim is fine if read narrowly.

Who gets value: people working on generation-based drug interaction prediction or molecule-to-text retrieval-augmented generation. It deserves a serious referee. I'd recommend conditional acceptance with a required pass on variance reporting and a toned-down inductive claim.

Referee Report

2 major / 5 minor

Summary. The paper proposes PKAG-DDI, a two-stage generative model for drug-drug interaction event (DDIE) text generation. Stage one is a pairwise knowledge selector (PKS) that takes the SMILES of two drugs, builds graph and fingerprint representations, exchanges information between the drugs via bidirectional cross-attention with a weight-reuse strategy, and predicts top-K biological functions for each drug. Stage two is a knowledge-augmented language model built on Galactica 1.3B with the MolTC graph-to-text adapter; it marginalizes over the K×K selected biological-function pairs and generates the DDIE text. Experiments on MecDDI and DDInter2.0 under random, cold-start, and scaffold splits compare against generation-based baselines (MolT5, MolTC) and classification baselines, and include ablations of the selector, the integration strategy, and the input modalities. The central claim is that PKAG-DDI outperforms existing methods for DDIE text generation, especially in inductive scenarios.

Significance. If the reported results hold, the paper makes a useful contribution: it introduces pairwise biological-function selection as a knowledge-augmentation mechanism for DDIE text generation, and it provides a concrete architecture with a public code and data release. The evaluation is broad—two datasets, three splits, generation and classification metrics, a gold-function upper bound, and several ablations—and the paper is candid about the limitation that the fixed knowledge set does not support zero-shot generalization to novel biological functions. The marginalization-based integration strategy is a reasonable response to the noise problem in retrieval-augmented generation. The main reservations concern statistical support for the headline inductive-scenario claim and the construction of the PKS training labels, both of which are addressable in revision.

major comments (2)

[Section 4.1, Table 1] The central claim that PKAG-DDI outperforms existing methods 'especially in challenging inductive scenarios' is not supported with statistical precision. All results are means of three runs, but no standard deviations, confidence intervals, or significance tests are reported anywhere in the paper. On the inductive rows where the claim is strongest, the differences are small and inconsistent: on MecDDI Cold Start, the METEOR lead over MolT5 is 64.34 versus 64.29 and ROUGE-L is 61.87 versus 61.95; on MecDDI Scaffold, PKAG-DDI trails MolT5 on METEOR (49.53 versus 50.29) and ROUGE-L (45.78 versus 46.45). Because the inductive scenarios are named in the abstract as the method's main strength, the manuscript should either add variance estimates and significance tests for all runs or explicitly temper the inductive-superiority claim.
[Section 3.2 (Training) and Appendix B] The gold biological-function labels for PKS training are selected by BM25 similarity between candidate biological functions and the target DDIE text. This means the first-stage supervision is derived from the same text that the second stage is trained to generate, creating a training-time leakage: the selector is rewarded for predicting functions that lexically overlap the target, and the PKS accuracies in Table 4 may therefore overstate how well the selector recovers biologically relevant functions without access to the target. At inference the target text is unavailable, so the end-to-end gains could shrink if the selector were trained on labels that do not depend on the target. Please add a robustness check, for example training PKS with alternative labels (all inherent functions, or a human-curated relevant function) and re-measuring end-to-end generation, or at least reporting agreement between BM25-chosen labels and independent annotations. This point does not invalidate the method, but it is load-bearing for the interpretation of the PKS and generation results.

minor comments (5)

[Section 3.3] In the model-architecture paragraph, the same symbol is used twice for the molecular token embeddings: 'Ta∈ RQ×dt and Ta∈ RQ×dt' should read 'Ta and Tb'.
[Table 1 caption] The caption says 'The abbreviations are BLEU-2, BLEU-4, and ROUGE-L' but the table also reports METEOR; please include METEOR in the list of abbreviations.
[Appendix D.1] The sentence 'The results shown in Figure 4 demonstrate that PKS outperforms the PKR w/ BoW and PKR w/ BERT' appears to refer to Table 4, not Figure 4; the cross-reference should be corrected.
[Appendix C.4] The paper recommends K=2 and states that larger K introduces more noise, but no sensitivity analysis for K is reported; a small K ∈ {1,2,3,4} ablation would make the choice of K and the noise claim concrete.
[Section 4.3 and Table 4] The efficiency comparison is reported only on MecDDI; since the inductive-superiority claim is a central theme, reporting PKS efficiency and accuracy on DDInter2.0 under the same splits would make the comparison more complete.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild target-derived supervision in PKS gold labels; the central generation claim remains empirically grounded and not circular.

self definitional [Section 3.2 (Training) and Appendix B]
"we use the BM25 (Robertson et al., 2009) to select the most similar pairwise biological function to the corresponding DDIE text as the gold biological function labels for training. ... the more token-similar the input and output text, the stronger the guidance of the input text for the label prediction, thereby improving the accuracy of prediction."

The gold biological function used to supervise PKS is defined by BM25 lexical similarity to the exact DDIE text that the generator must produce. Thus the intermediate 'knowledge' label is a deterministic compression of the target output, not an independent validated biological fact. PKS is trained to reproduce a target-derived label, and when PKS is accurate the LM is conditioned on a token-similar fragment of the reference. This makes the knowledge-selection stage partly self-referential. However, at inference PKS does not see the test text, so the reported PKAG-DDI generation scores are not forced by construction; the central claim is only mildly affected.

full rationale

The main derivation chain is not circular: PKS scores biological functions from molecular fingerprints and graphs with cross-attention, the generator marginalizes over top-K pairwise functions using Equation (10), and inference does not use the target text. The only identified circular element is the construction of PKS gold labels by BM25 similarity to the target DDIE text, described in Section 3.2 and justified in Appendix B. This makes the selector's supervision target-dependent, but it does not reduce the final generation claim to its inputs, since PKS accuracy at inference is imperfect (Table 4 shows A.@2 of 53.98 on Cold Start and 25.80 on Scaffold) and the generator is evaluated on unseen text. The upper-bound model PKAG-DDI* is explicitly an oracle using gold functions, so its near-perfect scores are presented as an upper bound rather than a claimed prediction. Self-citations to the authors' prior graph encoder and to MolTC pretrained parameters are component reuse, not load-bearing circularity. Statistical concerns about missing variance or significance tests are correctness risks, not circularity evidence. Overall, the paper has one mild target-derived supervision leak but its central generation claim is independent.

Assumptions & free parameters 2 free parameters · 2 assumptions · 0 invented entities

The method introduces no new biological or physical entities. Its reliance on external knowledge is captured by the MecDDI biological function set, and its main free parameters are K and lambda, both tuned on validation data. The most fragile assumption is the BM25-based label construction for PKS training, which the authors discuss but do not fully validate.

free parameters (2)

K (top-K biological functions) = 2
Number of candidate biological functions per drug; chosen as 2 because around 97% of drugs have fewer than three functions (Appendix C.4).
lambda (information flow in cross-attention) = tuned via Optuna, value not stated in text
Controls the residual information flow from the other drug in Equations 2 and 3; tuned as a hyperparameter.

assumptions (2)

domain assumption MecDDI biological function annotations are accurate and complete for both drugs in each DDI.
The method assumes that the MecDDI-provided functions correctly capture the relevant mechanisms; DDInter2.0 is filtered to include only pairs with such annotations (Section 4.1).
ad hoc to paper The BM25-selected gold biological function is the correct single-label supervision for training the selector.
Section 3.2 Training uses BM25 to pick the most similar function to the DDIE text; this heuristic is justified in Appendix B but is a modeling choice that can introduce label noise.

how reviews work

0 comments

Cite this review

Pith. "Pith review of PKAG-DDI: Pairwise Knowledge-Augmented Language Model for Drug-Drug Interaction Event Text Generation." pith.science (2026). https://pith.science/paper/FIN4M2WE

@misc{pith2026250719011,
  author       = {Pith},
  title        = {Pith review of: PKAG-DDI: Pairwise Knowledge-Augmented Language Model for Drug-Drug Interaction Event Text Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FIN4M2WE}},
  note         = {Machine review of arXiv:2507.19011}
}

read the original abstract

Drug-drug interactions (DDIs) arise when multiple drugs are administered concurrently. Accurately predicting the specific mechanisms underlying DDIs (named DDI events or DDIEs) is critical for the safe clinical use of drugs. DDIEs are typically represented as textual descriptions. However, most computational methods focus more on predicting the DDIE class label over generating human-readable natural language increasing clinicians' interpretation costs. Furthermore, current methods overlook the fact that each drug assumes distinct biological functions in a DDI, which, when used as input context, can enhance the understanding of the DDIE process and benefit DDIE generation by the language model (LM). In this work, we propose a novel pairwise knowledge-augmented generative method (termed PKAG-DDI) for DDIE text generation. It consists of a pairwise knowledge selector efficiently injecting structural information between drugs bidirectionally and simultaneously to select pairwise biological functions from the knowledge set, and a pairwise knowledge integration strategy that matches and integrates the selected biological functions into the LM. Experiments on two professional datasets show that PKAG-DDI outperforms existing methods in DDIE text generation, especially in challenging inductive scenarios, indicating its practicality and generalization.

Figures

Figures reproduced from arXiv: 2507.19011 by the authors.

**Figure 2.** The multi-class classification performance [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** The overall framework of PKAG-DDI. (a) is the overall framework of PKAG-DDI. It firstly uses a PKS [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Results of different integration strategies. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: The case study on DDInter2.0. Red text denotes content matching the reference labels. The underline [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

**Figure 6.** Figure 6: The ablation study about the information of [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

**Figure 7.** Figure 7: The variants of PKS. SMILES Molecular Graph Random Split Cold Start Split Scaffold Split ACC. ACC. ACC. ✓ × 94.94 43.95 21.55 × ✓ 94.95 43.86 19.98 ✓ ✓ 95.05 44.39 21.97 [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

**Figure 8.** Figure 8: The examples of predictions. The red text indicates matches, while the blue text indicates mismatches. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 18 canonical work pages

[1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Satanjeev Banerjee and Alon Lavie. 2005. Meteor: An automatic metric for mt evaluation with improved correlation with human judgments. In Proceedings of the acl workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, pages 65--72

2005
[4]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877--1901

2020
[5]

Yujie Chen, Tengfei Ma, Xixi Yang, Jianmin Wang, Bosheng Song, and Xiangxiang Zeng. 2021. MUFFIN : multi-scale feature fusion for drug–drug interaction prediction. Bioinformatics

work page 2021
[6]

Carl Edwards, Tuan Lai, Kevin Ros, Garrett Honke, Kyunghyun Cho, and Heng Ji. 2022. https://doi.org/10.18653/v1/2022.emnlp-main.26 Translation between molecules and natural language . In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 375--413, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics

doi:10.18653/v1/2022.emnlp-main.26 2022
[7]

Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. 2024. https://doi.org/10.1145/3637528.3671470 A survey on rag meeting llms: Towards retrieval-augmented large language models . In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD '24, page 6491–6501, New York, NY,...

arXiv 2024
[8]

Junfeng Fang, Shuai Zhang, Chang Wu, Zhengyi Yang, Zhiyuan Liu, Sihang Li, Kun Wang, Wenjie Du, and Xiang Wang. 2024. https://doi.org/10.18653/v1/2024.findings-acl.116 M ol TC : Towards molecular relational modeling in language models . In Findings of the Association for Computational Linguistics: ACL 2024, pages 1943--1958, Bangkok, Thailand. Association...

work page doi:10.18653/v1/2024.findings-acl.116 2024

Show all 40 references

[9]

Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, and Haofen Wang. 2023. Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997

2023 arXiv
[10]

Robert C Glen, Andreas Bender, Catrin H Arnby, Lars Carlsson, Scott Boyer, and James Smith. 2006. Circular fingerprints: flexible molecular descriptors with applications from physical chemistry to adme. IDrugs, 9(3):199

2006
[11]

Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Mingwei Chang. 2020. Retrieval augmented language model pre-training. In International conference on machine learning, pages 3929--3938. PMLR

2020
[12]

ZS Harris. 1954. Distributional structure

1954
[13]

Haohuai He, Guanxing Chen, and Calvin Yu-Chian Chen. 2022. 3dgt-ddi: 3d graph and text based neural network for drug--drug interaction prediction. Briefings in Bioinformatics, 23(3):bbac134

2022
[14]

Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low -rank adaptation of large language models. arXiv preprint arXiv:2106.09685

2021 arXiv
[15]

Wei Hu, Wei Zhang, Ying Zhou, Yongchao Luo, Xiuna Sun, Huimin Xu, Shuiyang Shi, Teng Li, Yichao Xu, Qianqian Yang, and others . 2023. MecDDI : Clarified Drug – Drug Interaction Mechanism Facilitating Rational Drug Use and Potential Drug – Drug Interaction Prediction . Journal ...

2023
[16]

Gautier Izacard, Patrick Lewis, Maria Lomeli, Lucas Hosseini, Fabio Petroni, Timo Schick, Jane Dwivedi-Yu, Armand Joulin, Sebastian Riedel, and Edouard Grave. 2024. Atlas: few-shot learning with retrieval augmented language models. J. Mach. Learn. Res., 24(1). Publisher: JMLR.org

2024
[17]

Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. https://doi.org/10.18653/v1/2020.emnlp-main.550 Dense passage retrieval for open-domain question answering . In Proceedings of the 2020 Conference on Empiric...

2020 doi
[18]

u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \

Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K \"u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \"a schel, et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Proc...

2020
[19]

Zimeng Li, Shichao Zhu, Bin Shao, Xiangxiang Zeng, Tong Wang, and Tie-Yan Liu. 2023. Dsn-ddi: an accurate and generalized framework for drug--drug interaction prediction by dual-view representation learning. Briefings in Bioinformatics, 24(1):bbac597

2023
[20]

Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74--81

2004
[21]

Weizhe Lin, Jinghong Chen, Jingbiao Mei, Alexandru Coca, and Bill Byrne. 2023 a . Fine-grained late-interaction multi-modal retrieval for retrieval augmented visual question answering. Advances in Neural Information Processing Systems, 36:22820--22840

2023
[22]

Xi Victoria Lin, Xilun Chen, Mingda Chen, Weijia Shi, Maria Lomeli, Rich James, Pedro Rodriguez, Jacob Kahn, Gergely Szilvasy, Mike Lewis, et al. 2023 b . Ra-dit: Retrieval-augmented dual instruction tuning. arXiv preprint arXiv:2310.01352

2023 arXiv
[23]

Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon, and Tie-Yan Liu. 2022. https://doi.org/10.1093/bib/bbac409 BioGPT : generative pre-trained transformer for biomedical text generation and mining . Briefings in Bioinformatics, 23(6)

2022 doi
[24]

Arnold K Nyamabo, Hui Yu, Zun Liu, and Jian-Yu Shi. 2022. Drug--drug interaction prediction with learnable size-adaptive molecular substructures. Briefings in Bioinformatics, 23(1):bbab441

2022
[25]

Arnold K Nyamabo, Hui Yu, and Jian-Yu Shi. 2021. https://doi.org/10.1093/bib/bbab133 SSI – DDI : substructure–substructure interactions for drug–drug interaction prediction . Briefings in Bioinformatics. \_eprint: https://academic.oup.com/bib/advance-article-pdf/doi/10.1093/bi...

2021 doi
[26]

Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311--318

2002
[27]

Ori Ram, Yoav Levine, Itay Dalmedigos, Dor Muhlgay, Amnon Shashua, Kevin Leyton-Brown, and Yoav Shoham. 2023. In-context retrieval-augmented language models. Transactions of the Association for Computational Linguistics, 11

2023
[28]

Stephen Robertson, Hugo Zaragoza, et al. 2009. The probabilistic relevance framework: Bm25 and beyond. Foundations and Trends in Information Retrieval , 3(4):333--389

2009
[29]

Jae Yong Ryu, Hyun Uk Kim, and Sang Yup Lee. 2018. Deep learning improves prediction of drug--drug and drug--food interactions. Proceedings of the national academy of sciences, 115(18):E4304--E4311

2018
[30]

Weijia Shi, Sewon Min, Michihiro Yasunaga, Minjoon Seo, Richard James, Mike Lewis, Luke Zettlemoyer, and Wen-tau Yih. 2024. Replug: Retrieval-augmented black-box language models. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computa...

2024
[31]

Devendra Singh, Siva Reddy, Will Hamilton, Chris Dyer, and Dani Yogatama. 2021. End-to-end training of multi-document reader and retriever for open-domain question answering. Advances in Neural Information Processing Systems, 34:25968--25981

2021
[32]

Ross Taylor, Marcin Kardas, Guillem Cucurull, Thomas Scialom, Anthony Hartshorn, Elvis Saravia, Andrew Poulton, Viktor Kerkez, and Robert Stojnic. 2022. Galactica: A large language model for science. arXiv preprint arXiv:2211.09085

2022 arXiv
[33]

Yao Tian, Jiacai Yi, Ningning Wang, Chengkun Wu, Jinfu Peng, Shao Liu, Guoping Yang, and Dongsheng Cao. 2024. https://doi.org/10.1093/nar/gkae726 DDInter 2.0: an enhanced drug interaction resource with expanded data coverage, new interaction types, and improved user interface ...

2024 doi
[34]

Yaqing Wang, Zaifei Yang, and Quanming Yao. 2024 a . Accurate and interpretable drug-drug interaction prediction enabled by knowledge subgraph learning. Communications Medicine, 4(1):59

2024
[35]

Ziyan Wang, Zhankun Xiong, Feng Huang, Xuan Liu, and Wen Zhang. 2024 b . Zeroddi: a zero-shot drug-drug interaction event prediction method with semantic enhanced learning and dual-modal uniform alignment. In Proceedings of the Thirty-Third International Joint Conference on Ar...

2024
[36]

Zhankun Xiong, Shichao Liu, Feng Huang, Ziyan Wang, Xuan Liu, Zhongfei Zhang, and Wen Zhang. 2023. Multi-relational contrastive learning graph neural network for drug-drug interaction event prediction. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37...

2023
[37]

Fangyuan Xu, Weijia Shi, and Eunsol Choi. 2024. https://openreview.net/forum?id=mlJLVigNHp RECOMP : Improving Retrieval - Augmented LMs with Context Compression and Selective Augmentation . In The Twelfth International Conference on Learning Representations

2024
[38]

Ziduo Yang, Weihe Zhong, Qiujie Lv, and Calvin Yu-Chian Chen. 2022. Learning size-adaptive molecular substructures for explainable drug–drug interaction prediction by substructure-aware graph neural network. Chemical science, 13(29):8693--8703. Publisher: Royal Society of Chemistry

2022
[39]

Yi Zhong, Gaozheng Li, Ji Yang, Houbing Zheng, Yongqiang Yu, Jiheng Zhang, Heng Luo, Biao Wang, and Zuquan Weng. 2024. Learning motif-based graphs for drug–drug interaction prediction via local–global self-attention. Nature Machine Intelligence, pages 1--12

2024
[40]

Xinyu Zhu, Yongliang Shen, and Weiming Lu. 2022. Molecular substructure-aware network for drug-drug interaction prediction. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pages 4757--4761

2022

Pith tools

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

[1] [1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

[2] [2]

write newline

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

[3] [3]

Satanjeev Banerjee and Alon Lavie. 2005. Meteor: An automatic metric for mt evaluation with improved correlation with human judgments. In Proceedings of the acl workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, pages 65--72

2005

[4] [4]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877--1901

2020

[5] [5]

Yujie Chen, Tengfei Ma, Xixi Yang, Jianmin Wang, Bosheng Song, and Xiangxiang Zeng. 2021. MUFFIN : multi-scale feature fusion for drug–drug interaction prediction. Bioinformatics

work page 2021

[6] [6]

Carl Edwards, Tuan Lai, Kevin Ros, Garrett Honke, Kyunghyun Cho, and Heng Ji. 2022. https://doi.org/10.18653/v1/2022.emnlp-main.26 Translation between molecules and natural language . In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 375--413, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics

doi:10.18653/v1/2022.emnlp-main.26 2022

[7] [7]

Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. 2024. https://doi.org/10.1145/3637528.3671470 A survey on rag meeting llms: Towards retrieval-augmented large language models . In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD '24, page 6491–6501, New York, NY,...

arXiv 2024

[8] [8]

Junfeng Fang, Shuai Zhang, Chang Wu, Zhengyi Yang, Zhiyuan Liu, Sihang Li, Kun Wang, Wenjie Du, and Xiang Wang. 2024. https://doi.org/10.18653/v1/2024.findings-acl.116 M ol TC : Towards molecular relational modeling in language models . In Findings of the Association for Computational Linguistics: ACL 2024, pages 1943--1958, Bangkok, Thailand. Association...

work page doi:10.18653/v1/2024.findings-acl.116 2024