{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T04:58:19Z","timestamp":1790657899873,"version":"4.1.0"},"reference-count":33,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T00:00:00Z","timestamp":1790640000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Res. Metr. Anal."],"abstract":"<jats:p>The rapid expansion of biomedical literature requires automated methods for accurate and efficient information extraction. This study addresses relation classification: given a pair of annotated biomedical entities in a research article title and abstract, assigning the relation that holds between them from a pre-defined set. We evaluate a retrieval-augmented generation (RAG) framework in which semantically similar annotated examples are retrieved from a training pool and supplied to a large language model as in-context demonstrations. Four open-source LLMs are compared against a supervised DeBERTa+CNN baseline on GutBrainIE, and four further corpora (ChemProt, BioRED, its BioCreative VIII release BC8, and DDI-2013) are used to test whether the findings generalize. We report macro- and micro-averaged precision, recall and F1-score, and the rate of predictions that fall outside the defined relation set. On GutBrainIE, Qwen3:14B with 25 retrieved examples reaches 0.877 macro and 0.866 micro F1, against 0.440 and 0.562 for the supervised baseline, while the smaller Llama-3.1-8B and Gemma3:12B models do not exceed the baseline on macro F1. Macro F1 rises from 0.105 without demonstrations to 0.877 with 25, with the steepest gains over the first ten retrieved examples and subsequent variation smaller than the run-to-run standard deviation. Predictions falling outside the defined relation set drop from 39 at zero shots to at most one between 20 and 35, so retrieval improves output validity as well as accuracy. Across the four further corpora these gains do not generalize uniformly: the best configuration exceeds the supervised baseline on ChemProt, improves macro but not micro F1 on BioRED and BC8, and falls below it on both measures on DDI-2013. Since DDI-2013 is the most balanced corpus in the set and, like ChemProt, is annotated within single sentences, that reversal is not explained by class imbalance or annotation scope, but is consistent with what the relation labels encode: whether they name relationships recoverable from biomedical knowledge, or boundaries set by an annotation guideline.<\/jats:p>","DOI":"10.3389\/frma.2026.1919972","type":"journal-article","created":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T04:33:14Z","timestamp":1790656394000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Biomedical retrieval-augmented generation for relation classification"],"prefix":"10.3389","volume":"11","author":[{"family":"Jannat","sequence":"first","affiliation":[{"name":"Department of Computer Science, Virginia Commonwealth University","place":["Richmond, VA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Charlie","family":"Dil","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Virginia Commonwealth University","place":["Richmond, VA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tom","family":"Arodz","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Virginia Commonwealth University","place":["Richmond, VA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bridget T.","family":"McInnes","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Virginia Commonwealth University","place":["Richmond, VA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,29]]},"reference":[{"key":"B1","article-title":"Qwen technical report","author":"Bai","year":"2023","journal-title":"arXiv"},{"key":"B2","author":"Baxter","year":"2010","journal-title":"Stockley's drug interactions pocket companion 2010"},{"key":"B3","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.wasp-main.6","author":"Brokman","year":"2025","journal-title":"A benchmark for end-to-end zero-shot biomedical relation extraction with LLMs: experiments with OpenAI models"},{"key":"B4","first-page":"1877","article-title":"\u201cLanguage models are few-shot learners,\u201d in: NIPS'20: proceedings of the 34th international conference on neural information processing systems","volume":"159","author":"Brown","year":"2020"},{"key":"B5","article-title":"Rq-rag: learning to refine queries for retrieval augmented generation","author":"Chan","year":"2024","journal-title":"arXiv"},{"key":"B6","doi-asserted-by":"publisher","first-page":"17754","DOI":"10.1609\/aaai.v38i16.29728","article-title":"Benchmarking large language models in retrieval-augmented generation. in","volume":"38","author":"Chen","year":"2024","journal-title":"Proceedings of the AAAI conference on artificial intelligence"},{"key":"B7","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N19-1423","author":"Devlin","year":"2019","journal-title":"Bert: pre-training of deep bidirectional transformers for language understanding. arXiv"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaf356","article-title":"Syract: zero-shot biomedical document-level relation extraction with synergistic rag and cot","author":"Dong","year":"2025","journal-title":"Bioinformatics"},{"key":"B9","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.naacl-main.194","article-title":"Re2g: Retrieve, rerank, generate","author":"Glass","year":"2022","journal-title":"arXiv"},{"key":"B10","article-title":"Retrieval-augmented generation with graphs (graphrag)","author":"Han","year":"2024","journal-title":"arXiv"},{"key":"B11","doi-asserted-by":"publisher","first-page":"914","DOI":"10.1016\/j.jbi.2013.07.011","article-title":"The DDI corpus: an annotated corpus with pharmacological substances and drug-drug interactions","volume":"46","author":"Herrero-Zazo","year":"2013","journal-title":"J. Biomed. Informat"},{"key":"B12","first-page":"16","article-title":"Information extraction","volume":"15","author":"Hobbs","year":"2010","journal-title":"Handbook Nat. Lang. Proc"},{"key":"B13","doi-asserted-by":"publisher","DOI":"10.1093\/database\/baae110","article-title":"The overview of the biored (biomedical relation extraction dataset) track at biocreative viii. Database:","author":"Islamaj","year":"2024"},{"key":"B14","unstructured":"2025"},{"key":"B15","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.emnlp-main.550","article-title":"Dense passage retrieval for open-domain question answering","author":"Karpukhin","year":"2020","journal-title":"EMNLP"},{"key":"B16","first-page":"141","article-title":"Overview of the BioCreative VI chemical-protein interaction track","author":"Krallinger","year":"2017","journal-title":"Proceedings of the BioCreative VI Workshop"},{"key":"B17","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.acl-long.43","article-title":"Hybgrag: Hybrid retrieval-augmented generation on textual and relational knowledge bases","author":"Lee","year":"2024","journal-title":"arXiv"},{"key":"B18","doi-asserted-by":"publisher","first-page":"9459","DOI":"10.48550\/arXiv.2005.11401","article-title":"Retrieval-augmented generation for knowledge-intensive NLP tasks","volume":"33","author":"Lewis","year":"2020","journal-title":"Adv. Neural Informat. Proc. Syst"},{"key":"B19","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.353","article-title":"Do you know what you are talking about? characterizing query-knowledge relevance for reliable retrieval augmented generation","author":"Li","year":"2024","journal-title":"arXiv"},{"key":"B20","doi-asserted-by":"publisher","first-page":"bbac282","DOI":"10.1093\/bib\/bbac282","article-title":"BioRED: a rich biomedical relation extraction dataset","volume":"23","author":"Luo","year":"2022","journal-title":"Briefings Bioinformat"},{"key":"B21","article-title":"Language models are few-shot learners","author":"Mann","year":"2020","journal-title":"arXiv"},{"key":"B22","volume-title":"Overview of gutbrainie@ clef 2025: gut-brain interplay information extraction","author":"Martinelli","year":"2025"},{"key":"B23","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2025.insights-1.11","article-title":"\u201cLlms are not zero-shot reasoners for biomedical information extraction,\u201d","volume-title":"Proceedings of the sixth workshop on insights from negative results in NLP","author":"Nagar","year":"2025"},{"key":"B24","doi-asserted-by":"publisher","DOI":"10.1145\/3731599.3767404","article-title":"Exploring distributed vector databases performance on hpc platforms: a study with qdrant","author":"Ockerman","year":"2025","journal-title":"arXiv"},{"key":"B25","article-title":"Relate: Relation extraction in biomedical abstracts with llms and ontology constraints","author":"Olasunkanmi","year":"2025","journal-title":"arXiv"},{"key":"B26","doi-asserted-by":"publisher","first-page":"baae104","DOI":"10.1093\/database\/baae104","article-title":"Optimized biomedical entity relation extraction method with data augmentation and classification using GPT-4 and Gemini","volume":"2024","author":"Phan","year":"2024","journal-title":"Database"},{"key":"B27","doi-asserted-by":"publisher","first-page":"333","DOI":"10.1561\/1500000019","article-title":"The probabilistic relevance framework : Bm25 and beyond","volume":"3","author":"Robertson","year":"2009","journal-title":"Foundat Trends registered Informat. Retrieval"},{"key":"B28","doi-asserted-by":"publisher","first-page":"566","DOI":"10.18653\/v1\/2022.findings-acl.48","article-title":"Extracting latent steering vectors from pretrained language models","volume":"2022","author":"Subramani","year":"2022","journal-title":"Findings of the association for computational linguistics: ACL"},{"key":"B29","article-title":"Dynamic parametric retrieval augmented generation for test-time knowledge enhancement","author":"Tan","year":"2025","journal-title":"arXiv"},{"key":"B30","article-title":"Gemma: Open models based on gemini research and technology","author":"Team","year":"2024","journal-title":"arXiv"},{"key":"B31","article-title":"Llama: open and efficient foundation language models","author":"Touvron","year":"2023","journal-title":"arXiv"},{"key":"B32","doi-asserted-by":"publisher","first-page":"4781","DOI":"10.1007\/s00521-021-06667-3","article-title":"Deep neural network-based relation extraction: an overview","volume":"34","author":"Wang","year":"2022","journal-title":"Neural Comput. Appl"},{"key":"B33","article-title":"Graphrag-bench: challenging domain-specific reasoning for evaluating graph retrieval-augmented generation","author":"Xiao","year":"2025","journal-title":"arXiv"}],"container-title":["Frontiers in Research Metrics and Analytics"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frma.2026.1919972\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T04:33:25Z","timestamp":1790656405000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frma.2026.1919972\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,29]]},"references-count":33,"alternative-id":["10.3389\/frma.2026.1919972"],"URL":"https:\/\/doi.org\/10.3389\/frma.2026.1919972","relation":{},"ISSN":["2504-0537"],"issn-type":[{"value":"2504-0537","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,29]]},"article-number":"1919972"}}