{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,6,15]],"date-time":"2026-06-15T09:54:03Z","timestamp":1781517243609,"version":"3.54.1"},"posted":{"date-parts":[[2026,6,11]]},"group-title":"Bioinformatics","reference-count":39,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T00:00:00Z","timestamp":1781136000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100009633","name":"NIH\/NICHD","doi-asserted-by":"crossref","award":["HD109765"],"award-info":[{"award-number":["HD109765"]}],"id":[{"id":"10.13039\/100009633","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/100000050","name":"National Heart Lung and Blood Institute","doi-asserted-by":"crossref","award":["U2CHL156291"],"award-info":[{"award-number":["U2CHL156291"]}],"id":[{"id":"10.13039\/100000050","id-type":"DOI","asserted-by":"crossref"}]},{"name":"HG\/NHGRI NIH HHS\/United States","award":["HG011898"],"award-info":[{"award-number":["HG011898"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,6,11]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>\n                  Down syndrome, caused by trisomy 21, increases the risk of diverse co-occurring conditions. With more than 34,000 related publications indexed in PubMed as of early 2026, keeping pace with this expanding literature is challenging. While general-purpose large language models are widely used for information retrieval, they often rely on broad training data rather than specific evidence. Retrieval-augmented generation (RAG) improves rigor and reliability of responses by linking model outputs to source texts. In research, source texts are peer-reviewed articles. Standard implementations treat all manuscript sections equally, allowing background text to rank as highly as experimental results. To focus model outputs on experimentally supported responses, we developed the T21 Research Assistant, a section-aware RAG system that prioritizes Results sections to ground responses in primary experimental evidence. The system draws exclusively from 1,789 open-access Down syndrome publications from PubMed Central, including 327 NIH INCLUDE-funded studies, and uses a multistage pipeline for query validation, retrieval, reranking, synthesis, and citation verification. Built on NVIDIA Nemotron models, it generates structured, cited responses. Evaluation using expert-curated questions demonstrated strong performance, achieving a BERTScore F1 of 0.712 and recall of 0.758, comparable to or exceeding leading proprietary and open-source models. T21 Research Assistant is available at:\n                  <jats:ext-link xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" ext-link-type=\"uri\" xlink:href=\"https:\/\/bioinformatics.cuanschutz.edu\/t21-res-assi\/\">https:\/\/bioinformatics.cuanschutz.edu\/t21-res-assi\/<\/jats:ext-link>\n                <\/jats:p>","DOI":"10.64898\/2026.06.08.730893","type":"posted-content","created":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T17:40:14Z","timestamp":1781199614000},"source":"Crossref","is-referenced-by-count":0,"title":["An AI-Powered Trisomy 21 Research Assistant"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0009-0001-5887-999X","authenticated-orcid":false,"given":"Sutanu","family":"Nandi","sequence":"first","affiliation":[{"name":"Department of Pharmacology, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zenitha","family":"Sundararajan","sequence":"additional","affiliation":[{"name":"Linda Crnic Institute for Down Syndrome, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marc","family":"Subirana-Gran\u00e9s","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9048-1941","authenticated-orcid":false,"given":"Joaquin M.","family":"Espinosa","sequence":"additional","affiliation":[{"name":"Department of Pharmacology, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"},{"name":"Linda Crnic Institute for Down Syndrome, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Milton","family":"Pividori","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kelly D.","family":"Sullivan","sequence":"additional","affiliation":[{"name":"Linda Crnic Institute for Down Syndrome, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"},{"name":"Department of Pediatrics, Section of Developmental Biology, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Matthew D.","family":"Galbraith","sequence":"additional","affiliation":[{"name":"Department of Pharmacology, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"},{"name":"Linda Crnic Institute for Down Syndrome, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3158-9682","authenticated-orcid":false,"given":"James C","family":"Costello","sequence":"additional","affiliation":[{"name":"Department of Pharmacology, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"},{"name":"Linda Crnic Institute for Down Syndrome, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"},{"name":"Department of Biomedical Informatics, University of Colorado Anschutz Medical Campus, Aurora, CO, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026061502150934000_2026.06.08.730893v1.1","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-026-03029-6"},{"key":"2026061502150934000_2026.06.08.730893v1.2","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pdig.0000877"},{"key":"2026061502150934000_2026.06.08.730893v1.3","unstructured":"Anthropic. 2026. \u201cClaude Sonnet 4.6.\u201d February 17. https:\/\/www.anthropic.com\/claude\/sonnet."},{"issue":"1","key":"2026061502150934000_2026.06.08.730893v1.4","doi-asserted-by":"crossref","first-page":"9","DOI":"10.1038\/s41572-019-0143-7","article-title":"Down Syndrome","volume":"6","year":"2020","journal-title":"Nature Reviews. Disease Primers"},{"issue":"5","key":"2026061502150934000_2026.06.08.730893v1.5","doi-asserted-by":"crossref","first-page":"e2022057010","DOI":"10.1542\/peds.2022-057010","article-title":"Health Supervision for Children and Adolescents with down Syndrome","volume":"149","year":"2022","journal-title":"Pediatrics"},{"key":"2026061502150934000_2026.06.08.730893v1.6","first-page":"1","article-title":"\u201cSimple and Effective Embedding Model for Single-Cell Biology Built from ChatGPT.\u201d Nature Biomedical Engineering","volume":"6","year":"2024","journal-title":"December"},{"key":"2026061502150934000_2026.06.08.730893v1.7","doi-asserted-by":"publisher","DOI":"10.1038\/nrn3314"},{"issue":"1","key":"2026061502150934000_2026.06.08.730893v1.8","doi-asserted-by":"crossref","first-page":"116835","DOI":"10.1016\/j.celrep.2025.116835","article-title":"Altered Hepatic Metabolism in Down Syndrome","volume":"45","year":"2026","journal-title":"Cell Reports"},{"key":"2026061502150934000_2026.06.08.730893v1.9","unstructured":"Face, Hugging. 2025. \u201cNvidia\/llama-Embed-Nemotron-8b \u00b7 Hugging Face.\u201d https:\/\/huggingface.co\/nvidia\/llama-embed-nemotron-8b."},{"key":"2026061502150934000_2026.06.08.730893v1.10","unstructured":"Face, Hugging. 2026. \u201cNvidia\/llama-Nemotron-Rerank-1b-v2 \u00b7 Hugging Face.\u201d https:\/\/huggingface.co\/nvidia\/llama-nemotron-rerank-1b-v2."},{"issue":"50","key":"2026061502150934000_2026.06.08.730893v1.11","doi-asserted-by":"crossref","first-page":"eadq3073","DOI":"10.1126\/sciadv.adq3073","article-title":"Integrated Analysis of Immunometabolic Interactions in Down Syndrome","volume":"10","year":"2024","journal-title":"Science Advances"},{"key":"2026061502150934000_2026.06.08.730893v1.12","unstructured":"\u201cGpt-oss:120b-Cloud.\u201d 2025. https:\/\/ollama.com\/library\/gpt-oss:120b-cloud."},{"key":"2026061502150934000_2026.06.08.730893v1.13","doi-asserted-by":"publisher","DOI":"10.1101\/2025.05.30.656746"},{"key":"2026061502150934000_2026.06.08.730893v1.14","unstructured":"\u201cLangChain: Observe, Evaluate, and Deploy Reliable AI Agents.\u201d 2022. https:\/\/www.langchain.com\/."},{"key":"2026061502150934000_2026.06.08.730893v1.15","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2005.11401"},{"key":"2026061502150934000_2026.06.08.730893v1.16","unstructured":"Lin, C-Y . 2004. \u201cROUGE: A Package for Automatic Evaluation of Summaries.\u201d In Text Summarisation Branches Out: Proceedings of the ACL-04 Workshop."},{"issue":"4","key":"2026061502150934000_2026.06.08.730893v1.17","doi-asserted-by":"crossref","first-page":"605","DOI":"10.1093\/jamia\/ocaf008","article-title":"Improving Large Language Model Applications in Biomedicine with Retrieval-Augmented Generation: A Systematic Review, Meta-Analysis, and Clinical Development Guidelines","volume":"32","year":"2025","journal-title":"Journal of the American Medical Informatics Association"},{"issue":"3","key":"2026061502150934000_2026.06.08.730893v1.18","doi-asserted-by":"crossref","first-page":"135","DOI":"10.1038\/s41582-018-0132-6","article-title":"Dementia in Down Syndrome: Unique Insights for Alzheimer Disease Research","volume":"15","year":"2019","journal-title":"Nature Reviews. Neurology"},{"key":"2026061502150934000_2026.06.08.730893v1.19","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-026-10265-5"},{"key":"2026061502150934000_2026.06.08.730893v1.20","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1802.03426"},{"issue":"9","key":"2026061502150934000_2026.06.08.730893v1.21","doi-asserted-by":"crossref","first-page":"611","DOI":"10.1097\/GIM.0b013e3181b2e34c","article-title":"The 50th Anniversary of the Discovery of Trisomy 21: The Past, Present, and Future of Research and Treatment of Down Syndrome","volume":"11","year":"2009","journal-title":"Genetics in Medicine: Official Journal of the American College of Medical Genetics"},{"issue":"9","key":"2026061502150934000_2026.06.08.730893v1.22","doi-asserted-by":"crossref","first-page":"e1005619","DOI":"10.1371\/journal.pcbi.1005619","article-title":"Ten Simple Rules for Structuring Papers","volume":"13","year":"2017","journal-title":"PLoS Computational Biology"},{"key":"2026061502150934000_2026.06.08.730893v1.23","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2509.06917"},{"key":"2026061502150934000_2026.06.08.730893v1.24","unstructured":"\u201cNCBI Entrez Programming Utilities (E-Utilities).\u201d 2009. https:\/\/www.ncbi.nlm.nih.gov\/pmc\/utils\/oa\/oa.fcgi."},{"key":"2026061502150934000_2026.06.08.730893v1.25","unstructured":"NVIDIA Corporation. 2026. \u201cNemotron-3-Super.\u201d https:\/\/ollama.com\/library\/nemotron-3-super."},{"key":"2026061502150934000_2026.06.08.730893v1.26","unstructured":"OpenAI. 2026. \u201cIntroducing GPT-5.4.\u201d OpenAI. https:\/\/openai.com\/index\/introducing-gpt-5-4\/."},{"key":"2026061502150934000_2026.06.08.730893v1.27","doi-asserted-by":"publisher","DOI":"10.3115\/1073083.1073135"},{"issue":"1","key":"2026061502150934000_2026.06.08.730893v1.28","doi-asserted-by":"crossref","first-page":"4766","DOI":"10.1038\/s41467-019-12739-9","article-title":"Trisomy 21 Activates the Kynurenine Pathway via Increased Dosage of Interferon Receptors","volume":"10","year":"2019","journal-title":"Nature Communications"},{"key":"2026061502150934000_2026.06.08.730893v1.29","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2205.01833"},{"key":"2026061502150934000_2026.06.08.730893v1.30","doi-asserted-by":"publisher","DOI":"10.1038\/s41551-025-01463-z"},{"key":"2026061502150934000_2026.06.08.730893v1.31","doi-asserted-by":"publisher","DOI":"10.7554\/elife.99323.3"},{"key":"2026061502150934000_2026.06.08.730893v1.32","doi-asserted-by":"publisher","DOI":"10.1038\/s41583-024-00866-2"},{"issue":"12","key":"2026061502150934000_2026.06.08.730893v1.33","doi-asserted-by":"crossref","first-page":"4495","DOI":"10.1007\/s00415-020-10179-w","article-title":"Neurologic Complications of Down Syndrome: A Systematic Review","volume":"268","year":"2021","journal-title":"Journal of Neurology"},{"key":"2026061502150934000_2026.06.08.730893v1.34","doi-asserted-by":"publisher","DOI":"10.7554\/eLife.16220"},{"key":"2026061502150934000_2026.06.08.730893v1.35","unstructured":"\u201cThe PubMed Central OAI-PMH API.\u201d 2001. https:\/\/pmc.ncbi.nlm.nih.gov\/tools\/oai\/."},{"issue":"10541","key":"2026061502150934000_2026.06.08.730893v1.36","doi-asserted-by":"crossref","first-page":"1779","DOI":"10.1016\/S0140-6736(26)00603-3","article-title":"Fabricated Citations: An Audit across 2\u00b75 Million Biomedical Papers","volume":"407","year":"2026","journal-title":"Lancet"},{"issue":"8","key":"2026061502150934000_2026.06.08.730893v1.37","doi-asserted-by":"crossref","first-page":"1677","DOI":"10.1038\/s41592-025-02748-6","article-title":"GeneAgent: Self-Verification Language Agent for Gene-Set Analysis Using Domain Databases","volume":"22","year":"2025","journal-title":"Nature Methods"},{"key":"2026061502150934000_2026.06.08.730893v1.38","doi-asserted-by":"publisher","DOI":"10.1007\/s00431-010-1253-0"},{"key":"2026061502150934000_2026.06.08.730893v1.39","unstructured":"Zhang, T. , V. Kishore , F. Wu , K. Q. Weinberger , and Y. Artzi . 2020. BERTScore: Evaluating Text Generation BERT. International Conference Learning Representations (ICLR 2020)."}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.06.08.730893","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,15]],"date-time":"2026-06-15T09:15:42Z","timestamp":1781514942000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.64898\/2026.06.08.730893"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,11]]},"references-count":39,"URL":"https:\/\/doi.org\/10.64898\/2026.06.08.730893","relation":{},"subject":[],"published":{"date-parts":[[2026,6,11]]},"subtype":"preprint"}}