{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T02:57:46Z","timestamp":1781665066067,"version":"3.54.5"},"posted":{"date-parts":[[2026,5,20]]},"group-title":"Hematology","reference-count":23,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,5,20]],"date-time":"2026-05-20T00:00:00Z","timestamp":1779235200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100001349","name":"National Medical Research Council","doi-asserted-by":"crossref","award":["MOH-001450-00"],"award-info":[{"award-number":["MOH-001450-00"]}],"id":[{"id":"10.13039\/501100001349","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,6,14]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:sec>\n                  <jats:title>Background<\/jats:title>\n                  <jats:p>\n                    Today, advancements in our understanding of cancer biology are increasingly attributed to large-scale clinical-molecular datasets. The case in point for multiple myeloma\u2013the second-most prevalent haematological malignancy\u2013is the CoMMpass study, a dataset with the paired clinical and sequencing data of 1,143 patients. However, the complexity of this rich dataset\u2014with 763 clinical parameters and summary data spread across &gt;20 files\u2014imposes hurdles to clinician-researchers interested in making simple queries like \u201cWhat percentage of patients relapse after VRD induction therapy?\u201d or \u201cCompare the overall survival of patients with high vs normal expression of\n                    <jats:italic>NSD2<\/jats:italic>\n                    \u201d.\n                  <\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Methods<\/jats:title>\n                  <jats:p>\n                    The rise of agentic AI over the past few years presents unparalleled opportunities to bridge this technical gap. We developed MyeGPT, an AI agent for clinical-molecular analysis of multiple myeloma. Based on the Reasoning-Acting (ReAct) framework, our agent converts natural language into\n                    <jats:italic>de novo<\/jats:italic>\n                    analyses grounded on the CoMMpass dataset, performs statistical analyses, and generates publication-quality plots. For validation, we created a benchmark of 20 calculation-intensive questions and designed two problems backed on published findings.\n                  <\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>MyeGPT achieves a mean reasoning-accuracy composite score of 79.4% on the internal benchmark and achieves inter-rater reliability of \u03ba = 0.965 with human bioinformaticians. It also reproduces published findings with near perfect accuracy. We deploy the agent as a ready-to-use browser application, enabling on-the-go hypothesis validation from a smartphone.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Conclusions<\/jats:title>\n                  <jats:p>MyeGPT demonstrates how agentic AI can eliminate the laborious scripting involved in analysing a large multi-omics dataset like CoMMpass. By increasing accessibility to a wide range of analyses from univariable statistics to transcriptome-wide hypothesis testing, MyeGPT can speed up clinical-cohort validation and hypothesis generation for multiple myeloma.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Key points<\/jats:title>\n                  <jats:list list-type=\"order\">\n                    <jats:list-item>\n                      <jats:p>We propose MyeGPT, a ReAct agent for the analysis and visualisation of multi-omics data of the CoMMpass study of Multiple Myeloma<\/jats:p>\n                    <\/jats:list-item>\n                    <jats:list-item>\n                      <jats:p>MyeGPT obtains a reasoning-accuracy composite score of 79.4 when evaluated on a numeric response question benchmark<\/jats:p>\n                    <\/jats:list-item>\n                    <jats:list-item>\n                      <jats:p>MyeGPT demonstrates high inter-rater reliability (Cohen\u2019s \u03ba 0.965) with human test takers on classifying functional high-risk patients<\/jats:p>\n                    <\/jats:list-item>\n                    <jats:list-item>\n                      <jats:p>We used MyeGPT to reproduce analyses in the official publication of CoMMpass release IA22 related to the PR RNA-seq subtype<\/jats:p>\n                    <\/jats:list-item>\n                    <jats:list-item>\n                      <jats:p>We applied MyeGPT on novel scenarios ranging from simple univariate queries, multivariate statistical testing, to transcriptome-wide multiple testing<\/jats:p>\n                    <\/jats:list-item>\n                  <\/jats:list>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Biographical note<\/jats:title>\n                  <jats:p>This study is a collaboration between researchers from the laboratory of Professor Chng Wee Joo, Senior Principal Investigator, Cancer Science Institute of Singapore and the Multiple Myeloma Research Foundation, USA.<\/jats:p>\n                <\/jats:sec>","DOI":"10.64898\/2026.05.14.26353252","type":"posted-content","created":{"date-parts":[[2026,5,20]],"date-time":"2026-05-20T22:35:13Z","timestamp":1779316513000},"source":"Crossref","is-referenced-by-count":0,"title":["\u201cMyeGPT: an AI agent for Multiple Myeloma\u201d"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0009-0006-0512-2877","authenticated-orcid":false,"given":"Jia Geng","family":"Chang","sequence":"first","affiliation":[{"name":"Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117597, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8943-5155","authenticated-orcid":false,"given":"Alexander M.","family":"Gout","sequence":"additional","affiliation":[{"name":"Multiple Myeloma Research Foundation, 383 Main Ave #7, Norwalk, CT 06851, United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jonathan","family":"Rodiger","sequence":"additional","affiliation":[{"name":"Multiple Myeloma Research Foundation, 383 Main Ave #7, Norwalk, CT 06851, United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1737-6322","authenticated-orcid":false,"given":"Tae-Hoon","family":"Chung","sequence":"additional","affiliation":[{"name":"Cancer Science Institute of Singapore, Singapore 117599, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-5084-9604","authenticated-orcid":false,"given":"George","family":"Mulligan","sequence":"additional","affiliation":[{"name":"Multiple Myeloma Research Foundation, 383 Main Ave #7, Norwalk, CT 06851, United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2578-8335","authenticated-orcid":false,"given":"Wee Joo","family":"Chng","sequence":"additional","affiliation":[{"name":"Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117597, Singapore"},{"name":"Cancer Science Institute of Singapore, Singapore 117599, Singapore"},{"name":"National University Cancer Institute, Singapore 119074, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026061619500971000_2026.05.14.26353252v5.1","doi-asserted-by":"crossref","unstructured":"Keats, J. J. et al. Interim Analysis Of The Mmrf Commpass Trial, a Longitudinal Study In Multiple Myeloma Relating Clinical Outcomes To Genomic and Immunophenotypic Profiles. Blood 122, 532 (2013).","DOI":"10.1182\/blood.V122.21.532.532"},{"key":"2026061619500971000_2026.05.14.26353252v5.2","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-024-01853-0"},{"key":"2026061619500971000_2026.05.14.26353252v5.3","doi-asserted-by":"crossref","first-page":"5618","DOI":"10.1182\/blood.V126.23.5618.5618","article-title":"Variations in Multiple Myeloma Disease Presentation By Race","volume":"126","year":"2015","journal-title":"Blood"},{"key":"2026061619500971000_2026.05.14.26353252v5.4","doi-asserted-by":"publisher","DOI":"10.1158\/1078-0432.CCR-21-0005"},{"key":"2026061619500971000_2026.05.14.26353252v5.5","doi-asserted-by":"crossref","unstructured":"Miller, C. et al. A Comparison of Clinical FISH and Sequencing Based FISH Estimates in Multiple Myeloma: An Mmrf Commpass Analysis. Blood 128, 374 (2016).","DOI":"10.1182\/blood.V128.22.374.374"},{"key":"2026061619500971000_2026.05.14.26353252v5.6","doi-asserted-by":"publisher","DOI":"10.1200\/JCO.2015.61.2267"},{"key":"2026061619500971000_2026.05.14.26353252v5.7","doi-asserted-by":"crossref","first-page":"285","DOI":"10.1016\/j.clml.2019.01.003","article-title":"Next Generation Sequencing-based Validation of the Revised International Staging System for Multiple Myeloma: An Analysis of the MMRF CoMMpass Study. Clinical Lymphoma","volume":"19","year":"2019","journal-title":"Myeloma and Leukemia"},{"key":"2026061619500971000_2026.05.14.26353252v5.8","doi-asserted-by":"crossref","first-page":"bbab050","DOI":"10.1093\/bib\/bbab050","article-title":"MMRFBiolinks: an R-package for integrating and analyzing MMRF-CoMMpass data","volume":"22","year":"2021","journal-title":"Brief Bioinform"},{"key":"2026061619500971000_2026.05.14.26353252v5.9","doi-asserted-by":"crossref","unstructured":"Settino, M. & Cannataro, M . MMRFVariant: Prioritizing variants in Multiple Myeloma. Informatics in Medicine Unlocked 39, 101271 (2023).","DOI":"10.1016\/j.imu.2023.101271"},{"key":"2026061619500971000_2026.05.14.26353252v5.10","unstructured":"Four AI Agent Strategies That Improve GPT-4 and GPT-3.5 Performance. Four AI Agent Strategies That Improve GPT-4 and GPT-3.5 Performance https:\/\/www.deeplearning.ai\/the-batch\/how-agents-can-improve-llm-performance\/ (2024)."},{"key":"2026061619500971000_2026.05.14.26353252v5.11","doi-asserted-by":"publisher","DOI":"10.1101\/2025.05.30.656746"},{"key":"2026061619500971000_2026.05.14.26353252v5.12","doi-asserted-by":"publisher","DOI":"10.1101\/2025.07.05.663295"},{"key":"2026061619500971000_2026.05.14.26353252v5.13","unstructured":"Superbio.ai. https:\/\/app.superbio.ai\/copilot."},{"key":"2026061619500971000_2026.05.14.26353252v5.14","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaf359"},{"key":"2026061619500971000_2026.05.14.26353252v5.15","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-019-0576-7"},{"key":"2026061619500971000_2026.05.14.26353252v5.16","doi-asserted-by":"crossref","unstructured":"Chang, J. G. , Chen, J. , Chew, G.-L. & Chng, W. J . MyeVAE: a multi-modal variational autoencoder for risk profiling of newly diagnosed multiple myeloma. BMC Artif . Intell. 1, 8 (2025).","DOI":"10.1186\/s44398-025-00009-2"},{"key":"2026061619500971000_2026.05.14.26353252v5.17","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2005.11401"},{"key":"2026061619500971000_2026.05.14.26353252v5.18","doi-asserted-by":"crossref","first-page":"1164","DOI":"10.1182\/blood.2024024429","article-title":"The ADAR1-regulated cytoplasmic dsRNA-sensing pathway is a novel mechanism of lenalidomide resistance in multiple myeloma","volume":"145","year":"2025","journal-title":"Blood"},{"key":"2026061619500971000_2026.05.14.26353252v5.19","doi-asserted-by":"publisher","DOI":"10.1038\/s41408-022-00631-7"},{"key":"2026061619500971000_2026.05.14.26353252v5.20","doi-asserted-by":"crossref","first-page":"224","DOI":"10.1038\/s43018-025-01072-4","article-title":"A single-cell atlas characterizes dysregulation of the bone marrow immune microenvironment associated with outcomes in multiple myeloma","volume":"7","year":"2026","journal-title":"Nat Cancer"},{"key":"2026061619500971000_2026.05.14.26353252v5.21","doi-asserted-by":"crossref","unstructured":"Behdenna, A. et al. pyComBat, a Python tool for batch effects correction in high-throughput molecular data using empirical Bayes methods. BMC Bioinformatics 24, 459 (2023).","DOI":"10.1186\/s12859-023-05578-5"},{"key":"2026061619500971000_2026.05.14.26353252v5.22","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2210.03629"},{"key":"2026061619500971000_2026.05.14.26353252v5.23","unstructured":"Chung, T.-H. , Chang, J. G. , Tan, T.-K. & Chng, W. J . Augmentation of the IMS\/IMWG high risk criteria for prognosis in multiple myeloma. Haematologica (Submitted for publication)."}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.05.14.26353252","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T02:50:24Z","timestamp":1781664624000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.64898\/2026.05.14.26353252"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,20]]},"references-count":23,"URL":"https:\/\/doi.org\/10.64898\/2026.05.14.26353252","relation":{},"subject":[],"published":{"date-parts":[[2026,5,20]]},"subtype":"preprint"}}