{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T20:50:25Z","timestamp":1792788625902,"version":"4.5.0"},"reference-count":38,"publisher":"Public Library of Science (PLoS)","issue":"10","license":[{"start":{"date-parts":[[2026,10,16]],"date-time":"2026-10-16T00:00:00Z","timestamp":1792108800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Innovation Fund for University Teachers, Gansu Provincial Education Department","award":["Grant no. 2025A-315 and 2025A-322"],"award-info":[{"award-number":["Grant no. 2025A-315 and 2025A-322"]}],"id":[{"id":"https:\/\/ror.org\/03w2z6524","id-type":"ROR","asserted-by":"crossref"}]},{"DOI":"10.13039\/501100004775","name":"Natural Science Foundation of Gansu Province","doi-asserted-by":"publisher","award":["Grant no. 24JRRA743"],"award-info":[{"award-number":["Grant no. 24JRRA743"]}],"id":[{"id":"10.13039\/501100004775","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Technology Application Innovation Program","award":["Grant no. 2024YY40"],"award-info":[{"award-number":["Grant no. 2024YY40"]}]}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"\n Medical multiple-choice question answering requires models to simultaneously understand question semantics, candidate-answer differences, and structured relations among medical entities in complex clinical contexts. To address the limitations that existing knowledge-enhanced methods often adopt question-level unified modeling and thus struggle to characterize local relational differences among different candidate options, while conventional low-rank adaptation lacks relation-pattern adaptivity, this paper proposes a knowledge-enhanced large language model framework that combines option-level knowledge subgraph representation fusion with relation-aware dynamic LoRA adaptation. The proposed method first constructs a local medical knowledge subgraph for the question stem and each candidate answer separately, and then injects structured knowledge into candidate representations through a graph encoder and gated fusion mechanism. Subsequently, multiple learned LoRA expert branches are dynamically combined through candidate-conditioned routing based on the relational information encoded in the enhanced candidate representations, enabling the model to obtain more suitable parameter update paths. The experimental results are reported as the mean and standard deviation over three random-seed runs. On the MedQA dataset, the proposed method achieves Accuracy scores of\n \n \n \n \n \n 62.90<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 0.82<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n and\n \n \n \n \n \n 62.84<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 1.04<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n , together with F1-Score values of\n \n \n \n \n \n 62.97<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 1.07<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n and\n \n \n \n \n \n 62.79<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 0.78<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n , using Llama3 8B and Qwen 7B, respectively. On the MedMCQA dataset, the corresponding Accuracy scores are\n \n \n \n \n \n 61.04<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 0.30<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n and\n \n \n \n \n \n 58.59<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 0.65<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n , while the F1-Score values reach\n \n \n \n \n \n 61.21<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 0.54<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n and\n \n \n \n \n \n 57.77<\/mml:mn>\n %<\/mml:mi>\n \u00b1<\/mml:mo>\n 0.57<\/mml:mn>\n %<\/mml:mi>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n , respectively. Experimental results show that the proposed method can effectively improve knowledge understanding, candidate discrimination, and reasoning stability in medical multiple-choice question answering.\n <\/jats:p>","DOI":"10.1371\/journal.pone.0349628","type":"journal-article","created":{"date-parts":[[2026,10,16]],"date-time":"2026-10-16T17:37:32Z","timestamp":1792172252000},"page":"e0349628","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Relation-aware dynamic knowledge subgraph enhanced large language models for medical multiple-choice question answering"],"prefix":"10.1371","volume":"21","author":[{"given":"Yu","family":"Ma","sequence":"first","affiliation":[{"name":"Gansu Police College","place":["Lanzhou, China"]},{"name":"Police Embodied Intelligence and Unmanned Systems Innovation Center","place":["Lanzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tianning","family":"Di","sequence":"additional","affiliation":[{"name":"The Second Hospital of Lanzhou University","place":["Lanzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiangpeng","family":"Xu","sequence":"additional","affiliation":[{"name":"Gansu Provincial Public Security Department","place":["Lanzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-6404-5937","authenticated-orcid":true,"given":"Yue","family":"Wu","sequence":"additional","affiliation":[{"name":"Gansu Police College","place":["Lanzhou, China"]},{"name":"Police Embodied Intelligence and Unmanned Systems Innovation Center","place":["Lanzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,10,16]]},"reference":[{"issue":"7972","key":"pone.0349628.ref001","doi-asserted-by":"crossref","first-page":"172","DOI":"10.1038\/s41586-023-06291-2","article-title":"Large language models encode clinical knowledge","volume":"620","author":"K Singhal","year":"2023","journal-title":"Nature"},{"issue":"3","key":"pone.0349628.ref002","doi-asserted-by":"crossref","first-page":"943","DOI":"10.1038\/s41591-024-03423-7","article-title":"Toward expert-level medical question answering with large language models","volume":"31","author":"K Singhal","year":"2025","journal-title":"Nat Med"},{"key":"pone.0349628.ref003","unstructured":"Zhou H, Liu F, Gu B, Zou X, Huang J, Wu J. A survey of large language models in medicine: Progress, application, and challenge. 2023. https:\/\/arxiv.org\/abs\/2311.05112"},{"key":"pone.0349628.ref004","doi-asserted-by":"crossref","unstructured":"Yasunaga M, Ren H, Bosselut A, Liang P, Leskovec J. QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering. In: Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021. 535\u201346. https:\/\/aclanthology.org\/2021.naacl-main.45\/","DOI":"10.18653\/v1\/2021.naacl-main.45"},{"key":"pone.0349628.ref005","unstructured":"Zhang X, Bosselut A, Yasunaga M, Ren H, Liang P, Manning CD. GreaseLM: Graph REASoning Enhanced Language Models for Question Answering. In: 2022."},{"key":"pone.0349628.ref006","doi-asserted-by":"crossref","unstructured":"Yasunaga M, Bosselut A, Ren H, Zhang X, Manning CD, Liang P, et al. Deep Bidirectional Language-Knowledge Graph Pretraining. In: Advances in Neural Information Processing Systems 35, 2022. 37309\u201323. https:\/\/doi.org\/10.52202\/068431-2704","DOI":"10.52202\/068431-2704"},{"issue":"2","key":"pone.0349628.ref007","doi-asserted-by":"crossref","first-page":"1015","DOI":"10.1007\/s13042-024-02318-w","article-title":"Large language models for medicine: a survey","volume":"16","author":"Y Zheng","year":"2024","journal-title":"Int J Mach Learn & Cyber"},{"issue":"3","key":"pone.0349628.ref008","doi-asserted-by":"crossref","first-page":"100943","DOI":"10.1016\/j.patter.2024.100943","article-title":"Can large language models reason about medical questions?","volume":"5","author":"V Li\u00e9vin","year":"2024","journal-title":"Patterns (N Y)"},{"key":"pone.0349628.ref009","doi-asserted-by":"crossref","first-page":"104861","DOI":"10.1016\/j.jbi.2025.104861","article-title":"A review on knowledge graphs for healthcare: Resources, applications, and promises","volume":"169","author":"H Cui","year":"2025","journal-title":"J Biomed Inform"},{"key":"pone.0349628.ref010","unstructured":"Liu Z, Peng E, Yan S, Li G, Hao T. T-know: a knowledge graph-based question answering and information retrieval system for traditional Chinese medicine. In: Proceedings of the 27th International Conference on Computational Linguistics: System Demonstrations, 2018. 15\u20139."},{"key":"pone.0349628.ref011","doi-asserted-by":"crossref","unstructured":"Chen J, Zhou J, Shi Z, Fan B, Luo C. Knowledge Abstraction Matching for Medical Question Answering. In: 2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2019. 342\u20137. https:\/\/doi.org\/10.1109\/bibm47256.2019.8982973","DOI":"10.1109\/BIBM47256.2019.8982973"},{"key":"pone.0349628.ref012","doi-asserted-by":"crossref","unstructured":"Yang R, Liu H, Marrese-Taylor E, Zeng Q, Ke Y, Li W, et al. KG-Rank: Enhancing Large Language Models for Medical QA with Knowledge Graphs and Ranking Techniques. In: Proceedings of the 23rd Workshop on Biomedical Natural Language Processing, 2024. 155\u201366. https:\/\/doi.org\/10.18653\/v1\/2024.bionlp-1.13","DOI":"10.18653\/v1\/2024.bionlp-1.13"},{"key":"pone.0349628.ref013","unstructured":"Cabello L, Martin-Turrero C, Akujuobi U, S\u00f8gaard A, Bobed C. Meg: Medical knowledge-augmented large language models for question answering. 2024. https:\/\/arxiv.org\/abs\/241103883"},{"issue":"2","key":"pone.0349628.ref014","doi-asserted-by":"crossref","first-page":"2172","DOI":"10.1007\/s10489-024-05282-8","article-title":"Are my answers medically accurate? Exploiting medical knowledge graphs for medical question answering","volume":"54","author":"A Zafar","year":"2024","journal-title":"Appl Intell"},{"key":"pone.0349628.ref015","doi-asserted-by":"crossref","first-page":"112837","DOI":"10.1016\/j.knosys.2024.112837","article-title":"KoSEL: Knowledge subgraph enhanced large language model for medical question answering","volume":"309","author":"Z Zeng","year":"2025","journal-title":"Knowledge-Based Systems"},{"issue":"7","key":"pone.0349628.ref016","doi-asserted-by":"crossref","first-page":"5317","DOI":"10.1109\/JBHI.2025.3547444","article-title":"Infusing Multi-Hop Medical Knowledge Into Smaller Language Models for Biomedical Question Answering","volume":"29","author":"J Chen","year":"2025","journal-title":"IEEE J Biomed Health Inform"},{"key":"pone.0349628.ref017","unstructured":"Zafar A, Mishra K, Ekbal A. Medex: Enhancing medical question-answering with first-order logic based reasoning and knowledge injection. In: Proceedings of the 31st International Conference on Computational Linguistics, 2025. 9701\u201320."},{"key":"pone.0349628.ref018","doi-asserted-by":"crossref","unstructured":"Rezaei MR, Fard RS, Parker JL, Krishnan RG, Lankarany M. Agentic medical knowledge graphs enhance medical question answering: bridging the gap between llms and evolving medical knowledge. 2025. https:\/\/arxiv.org\/abs\/250213010","DOI":"10.18653\/v1\/2025.findings-emnlp.679"},{"key":"pone.0349628.ref019","unstructured":"Li F, Chen Y, Liu H, Yang R, Yuan H, Jiang Y. Mkg-rank: Enhancing large language models with knowledge graph for multilingual medical question answering. 2025. https:\/\/arxiv.org\/abs\/250316131"},{"issue":"2","key":"pone.0349628.ref020","first-page":"3","article-title":"Lora: Low-rank adaptation of large language models","volume":"1","author":"EJ Hu","year":"2022","journal-title":"ICLR"},{"key":"pone.0349628.ref021","unstructured":"Zhang Q, Chen M, Bukharin A, Karampatziakis N, He P, Cheng Y. Adalora: Adaptive budget allocation for parameter-efficient fine-tuning. 2023. https:\/\/arxiv.org\/abs\/2303.10512"},{"key":"pone.0349628.ref022","unstructured":"Luo T, Lei J, Lei F, Liu W, He S, Zhao J. Moelora: Contrastive learning guided mixture of experts on parameter-efficient fine-tuning for large language models. 2024. https:\/\/arxiv.org\/abs\/240212851"},{"key":"pone.0349628.ref023","unstructured":"Zhao Z, Gan L, Wang G, Hu Y, Shen T, Yang H. Retrieval-augmented mixture of lora experts for uploadable machine learning. In: 2024. https:\/\/arxiv.org\/abs\/2406.16989"},{"key":"pone.0349628.ref024","unstructured":"Yang Y, Zeng A, Yang X. Towards specialized generalists: A multi-task MoE-LoRA framework for domain-specific LLM adaptation. 2026. https:\/\/arxiv.org\/abs\/260107935"},{"key":"pone.0349628.ref025","doi-asserted-by":"crossref","unstructured":"Zhang D, Zhang K, Chu S, Wu L, Li X, Wei S. MoRE: A Mixture of Low-Rank Experts for Adaptive Multi-Task Learning. In: Findings of the Association for Computational Linguistics: ACL 2025, 2025. 1311\u201324. https:\/\/doi.org\/10.18653\/v1\/2025.findings-acl.68","DOI":"10.18653\/v1\/2025.findings-acl.68"},{"key":"pone.0349628.ref026","doi-asserted-by":"crossref","unstructured":"Lyu Y, Yan L, Wang S, Shi H, Yin D, Ren P, et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. In: Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 2024. 14535\u201356. https:\/\/doi.org\/10.18653\/v1\/2024.emnlp-main.805","DOI":"10.18653\/v1\/2024.emnlp-main.805"},{"key":"pone.0349628.ref027","doi-asserted-by":"crossref","first-page":"1468","DOI":"10.1016\/j.procs.2025.09.268","article-title":"Efficient medical question answering through QLoRA fine-tuning and knowledge distillation","volume":"270","author":"B Brzek","year":"2025","journal-title":"Procedia Computer Science"},{"key":"pone.0349628.ref028","unstructured":"He J, Li P, Liu G, He G, Chen Z, Zhong S. PeFoMed: Parameter efficient fine-tuning of multimodal large language models for medical imaging. 2024. https:\/\/arxiv.org\/abs\/2401.02797"},{"key":"pone.0349628.ref029","doi-asserted-by":"crossref","first-page":"270","DOI":"10.1016\/j.eng.2025.02.010","article-title":"Toward a Large Language Model-Driven Medical Knowledge Retrieval and QA System: Framework Design and Evaluation","volume":"50","author":"Y Liu","year":"2025","journal-title":"Engineering"},{"issue":"14","key":"pone.0349628.ref030","doi-asserted-by":"crossref","first-page":"6421","DOI":"10.3390\/app11146421","article-title":"What Disease Does This Patient Have? A Large-Scale Open Domain Question Answering Dataset from Medical Exams","volume":"11","author":"D Jin","year":"2021","journal-title":"Applied Sciences"},{"key":"pone.0349628.ref031","unstructured":"Pal A, Umapathi LK, Sankarasubbu M. Medmcqa: A large-scale multi-subject multi-choice dataset for medical domain question answering. In: 2022. 248\u201360."},{"key":"pone.0349628.ref032","doi-asserted-by":"crossref","unstructured":"Wang Z, Shen Z, He Y, Sun G, Wang H, Lyu L, et al. FLoRA: Federated Fine-Tuning Large Language Models with Heterogeneous Low-Rank Adaptations. In: Advances in Neural Information Processing Systems 37, 2024. 22513\u201333. https:\/\/doi.org\/10.52202\/079017-0708","DOI":"10.52202\/079017-0708"},{"key":"pone.0349628.ref033","unstructured":"Yang Z, Qian J, Zhang H, Tan K, Huang YA, Huang ZA, et al. Med-REFL: Enhancing Complex Reasoning via Fine-grained Self-Correction."},{"key":"pone.0349628.ref034","unstructured":"Li Y, Yu Y, Liang C, He P, Karampatziakis N, Chen W. Loftq: Lora-fine-tuning-aware quantization for large language models. 2023. https:\/\/arxiv.org\/abs\/231008659"},{"key":"pone.0349628.ref035","doi-asserted-by":"crossref","unstructured":"Tian C, Shi Z, Guo Z, Li L, Xu C. HydraLoRA: An Asymmetric LoRA Architecture for Efficient Fine-Tuning. In: Advances in Neural Information Processing Systems 37, 2024. 9565\u201384. https:\/\/doi.org\/10.52202\/079017-0304","DOI":"10.52202\/079017-0304"},{"key":"pone.0349628.ref036","doi-asserted-by":"crossref","unstructured":"Bian J, Wang L, Zhang L, Xu J. LoRA-FAIR: Federated LoRA Fine-Tuning with Aggregation and Initialization Refinement. In: 2025 IEEE\/CVF International Conference on Computer Vision (ICCV), 2025. 3737\u201346. https:\/\/doi.org\/10.1109\/iccv51701.2025.00356","DOI":"10.1109\/ICCV51701.2025.00356"},{"key":"pone.0349628.ref037","unstructured":"Su X, Wang Y, Gao S, Liu X, Giunchiglia V, Clevert DA. KGARevion: An AI Agent for Knowledge-Intensive Biomedical QA. In: International Conference on Learning Representations. 2025."},{"issue":"1","key":"pone.0349628.ref038","doi-asserted-by":"crossref","first-page":"240","DOI":"10.1038\/s41746-025-01653-8","article-title":"Small Language Models Learn Enhanced Reasoning Skills from Medical Textbooks","volume":"8","author":"H Kim","year":"2025","journal-title":"npj Digital Medicine"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0349628","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,16]],"date-time":"2026-10-16T17:37:38Z","timestamp":1792172258000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0349628"}},"subtitle":[],"editor":[{"given":"Ivan","family":"Zyrianoff","sequence":"first","affiliation":[{"name":"Alma Mater Studiorum Universita di Bologna: Universita degli Studi di Bologna","place":["ITALY"]}],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,10,16]]},"references-count":38,"journal-issue":{"issue":"10","published-online":{"date-parts":[[2026,10,16]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0349628","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,16]]}}}