{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T23:45:12Z","timestamp":1792799112288,"version":"4.5.0"},"reference-count":35,"publisher":"MDPI AG","issue":"4","license":[{"start":{"date-parts":[[2026,4,18]],"date-time":"2026-04-18T00:00:00Z","timestamp":1776470400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100003725","name":"the National Research Foundation of Korea","doi-asserted-by":"crossref","award":["RS-2023-NR076559"],"award-info":[{"award-number":["RS-2023-NR076559"]}],"id":[{"id":"10.13039\/501100003725","id-type":"DOI","asserted-by":"crossref"}]},{"award":["RS-2023-NR076559"],"award-info":[{"award-number":["RS-2023-NR076559"]}],"id":[{"id":"https:\/\/ror.org\/013aysd81","id-type":"ROR","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Pharmaceutics"],"abstract":"<jats:p>Background\/Objectives: Population pharmacokinetic (PopPK) modeling in NONMEM requires iterative, expertise-dependent workflows. Na\u00efve zero-shot prompting of general-purpose large language models (LLMs) typically produces NONMEM code that fails to execute. This study introduces PKGPT, a recursive agentic LLM system designed to automate NONMEM-based PopPK model development and benchmarks its performance against human expert models. Methods: PKGPT, powered by Google\u2019s Gemini 3.0 Flash, embeds pharmacometrics expertise into phase-specific expert-agent prompts orchestrated across five sequential phases: base model establishment, structural diagnostics, overfitting reduction, random-effects optimization, and covariate analysis. The system recursively executes NONMEM, parses outputs, and iteratively refines control streams. PKGPT was evaluated on three public datasets (warfarin, theophylline, and tobramycin) and benchmarked against independently developed human expert models. Results: PKGPT consistently produced executable, converging NONMEM models across all three datasets. In warfarin, both PKGPT and the human expert selected a one-compartment oral structure (ADVAN2), but the expert achieved a lower OFV (294.41 vs. 484.43) via covariate scaling. In theophylline, PKGPT produced parameter estimates close to the expert solution (Ka = 1.59 vs. 1.46 h\u22121; CL = 0.0399 vs. 0.0404 L\/h\/kg). In tobramycin, PKGPT correctly identified a two-compartment structure but produced physiologically implausible peripheral volume estimates (V2 = 149 L vs. expert\u2019s 13.2 L). Across datasets, PKGPT did not identify clinically established covariates, and run-to-run reproducibility was variable. Conclusions: PKGPT substantially improves the robustness and usability of LLM-generated NONMEM code compared with na\u00efve zero-shot prompting, accelerating model drafting and iterative refinement, but physiological plausibility and clinical interpretability still require a human-in-the-loop oversight.<\/jats:p>","DOI":"10.3390\/pharmaceutics18040501","type":"journal-article","created":{"date-parts":[[2026,4,20]],"date-time":"2026-04-20T09:37:42Z","timestamp":1776677862000},"page":"501","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":2,"title":["PKGPT: Expert-Orchestrated Recursive LLM Agent for Automated NONMEM PopPK Modeling with Human Benchmarking"],"prefix":"10.3390","volume":"18","author":[{"given":"Hoyoung","family":"Kwack","sequence":"first","affiliation":[{"name":"Department of Pharmacy and Yonsei Institute of Pharmaceutical Sciences, Yonsei University, Incheon 21983, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7681-7429","authenticated-orcid":false,"given":"Hyunseung","family":"Kong","sequence":"additional","affiliation":[{"name":"Interdisciplinary Program in Bioinformatics, Seoul National University, Seoul 08826, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-0293-5326","authenticated-orcid":false,"given":"Jiwoo","family":"Lim","sequence":"additional","affiliation":[{"name":"Department of Integrative Biotechnology, College of Pharmacy, Yonsei University, Incheon 21983, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9890-0389","authenticated-orcid":false,"given":"Byoung-Tak","family":"Zhang","sequence":"additional","affiliation":[{"name":"Interdisciplinary Program in Bioinformatics, Seoul National University, Seoul 08826, Republic of Korea"},{"name":"Department of Computer Science, Seoul National University, Seoul 08826, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7841-8349","authenticated-orcid":false,"given":"Jongsung","family":"Hahn","sequence":"additional","affiliation":[{"name":"College of Pharmacy, Jeonbuk National University, Jeonju 54896, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8408-5907","authenticated-orcid":false,"given":"Min Jung","family":"Chang","sequence":"additional","affiliation":[{"name":"Department of Pharmacy and Yonsei Institute of Pharmaceutical Sciences, Yonsei University, Incheon 21983, Republic of Korea"},{"name":"Department of Integrative Biotechnology, College of Pharmacy, Yonsei University, Incheon 21983, Republic of Korea"},{"name":"Graduate Program of Industrial Pharmaceutical Science, Yonsei University, Incheon 21983, Republic of Korea"},{"name":"Department of Pharmaceutical Medicine and Regulatory Science, Yonsei University, Incheon 21983, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,4,18]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"1127","DOI":"10.1002\/psp4.12699","article-title":"Perspective on Model-Informed Drug Development","volume":"10","author":"Lesko","year":"2021","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"899","DOI":"10.1002\/cpt.1363","article-title":"Model-Informed Drug Development: Current US Regulatory Practice and Future Considerations","volume":"105","author":"Wang","year":"2019","journal-title":"Clin. Pharmacol. Ther."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"627","DOI":"10.2165\/11593210-000000000-00000","article-title":"Impact of Pharmacometric Analyses on New Drug Approval and Labelling Decisions: A Review of 198 Submissions between 2000 and 2008","volume":"50","author":"Lee","year":"2011","journal-title":"Clin. Pharmacokinet."},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"1697","DOI":"10.1007\/s40262-022-01182-7","article-title":"A Survey of Population Pharmacokinetic Reports Submitted to the USFDA: An Analysis of Common Issues in NDA and BLA from 2012 to 2021","volume":"61","author":"Hsu","year":"2022","journal-title":"Clin. Pharmacokinet."},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"E503","DOI":"10.1208\/aapsj070351","article-title":"Impact of Pharmacometrics on Drug Approval and Labeling Decisions: A Survey of 42 New Drug Applications","volume":"7","author":"Bhattaram","year":"2005","journal-title":"AAPS J."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"65","DOI":"10.1002\/cpt.1923","article-title":"Drug Dosing Recommendations for All Patients: A Roadmap for Change","volume":"109","author":"Powell","year":"2021","journal-title":"Clin. Pharmacol. Ther."},{"key":"ref_7","unstructured":"U.S. Food and Drug Administration (2026, February 22). Population Pharmacokinetics: Guidance for Industry, Available online: https:\/\/www.fda.gov\/media\/128793\/download."},{"key":"ref_8","unstructured":"European Medicines Agency (2026, February 22). Guideline on Reporting the Results of Population Pharmacokinetic Analyses (CHMP\/EWP\/185990\/06). Available online: https:\/\/www.ema.europa.eu\/en\/reporting-results-population-pharmacokinetic-analyses-scientific-guideline."},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"93","DOI":"10.1002\/psp4.12049","article-title":"Good Practices in Model-Informed Drug Discovery and Development: Practice, Application, and Documentation","volume":"5","author":"Marshall","year":"2016","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"418","DOI":"10.1002\/psp4.12196","article-title":"Advanced Methods for Dose and Regimen Finding during Drug Development: Summary of the EMA\/EFPIA Workshop on Dose Finding (London 4\u20135 December 2014)","volume":"6","author":"Manolis","year":"2017","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"1073","DOI":"10.1002\/jcph.579","article-title":"A Proposal for Scientific Framework Enabling Specific Population Drug Dosing Recommendations","volume":"55","author":"Jadhav","year":"2015","journal-title":"J. Clin. Pharmacol."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"e6","DOI":"10.1038\/psp.2012.4","article-title":"Basic Concepts in Population Modeling, Simulation, and Model-Based Drug Development","volume":"1","author":"Mould","year":"2012","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"1702","DOI":"10.1345\/aph.1D374","article-title":"Population Pharmacokinetics I: Background, Concepts, and Models","volume":"38","author":"Ette","year":"2004","journal-title":"Ann. Pharmacother."},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"e51","DOI":"10.1038\/psp.2013.26","article-title":"Establishing Best Practices and Guidance in Population Modeling: An Experience with an Internal Population Pharmacokinetic Analysis Guidance","volume":"2","author":"Byon","year":"2013","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"1210","DOI":"10.1002\/psp4.12838","article-title":"Efficient and Relevant Stepwise Covariate Model Building for Pharmacometrics","volume":"11","author":"Svensson","year":"2022","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"710","DOI":"10.1002\/psp4.13115","article-title":"Covariate Modeling in Pharmacometrics: General Points for Consideration","volume":"13","author":"Maas","year":"2024","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"e50","DOI":"10.1038\/psp.2013.24","article-title":"Modeling and Simulation Workbench for NONMEM: Tutorial on Pirana, PsN, and Xpose","volume":"2","author":"Keizer","year":"2013","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_18","unstructured":"Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H.P.O., Kaplan, J., Edwards, H., Burda, Y., Joseph, N., and Brockman, G. (2021). Evaluating Large Language Models Trained on Code. arXiv."},{"key":"ref_19","unstructured":"Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., and Cao, Y. (2023, January 1\u20135). ReAct: Synergizing Reasoning and Acting in Language Models. Proceedings of the International Conference on Learning Representations (ICLR), Kigali, Rwanda."},{"key":"ref_20","unstructured":"Wu, Q., Bansal, G., Zhang, J., Wu, Y., Li, B., Zhu, E., Jiang, L., Zhang, X., Zhang, S., and Liu, J. (2023). AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation. arXiv."},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Shinn, N., Cassano, F., Berman, E., Gopinath, A., Narasimhan, K., and Yao, S. (2023, January 10\u201316). Reflexion: Language Agents with Verbal Reinforcement Learning. Proceedings of the Advances in Neural Information Processing Systems (NeurIPS), New Orleans, LA, USA.","DOI":"10.52202\/075280-0377"},{"key":"ref_22","unstructured":"Olausson, T.X., Inala, J.P., Wang, C., Gao, J., and Solar-Lezama, A. (2024, January 7\u201311). Is Self-Repair a Silver Bullet for Code Generation?. Proceedings of the International Conference on Learning Representations (ICLR), Vienna, Austria."},{"key":"ref_23","doi-asserted-by":"crossref","unstructured":"Yang, J., Jimenez, C.E., Wettig, A., Lieret, K., Yao, S., Narasimhan, K., and Press, O. (2024, January 9\u201315). SWE-Agent: Agent-Computer Interfaces Enable Automated Software Engineering. Proceedings of the Advances in Neural Information Processing Systems (NeurIPS), Vancouver, BC, Canada.","DOI":"10.52202\/079017-1601"},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"773","DOI":"10.1080\/01621459.1995.10476572","article-title":"Bayes Factors","volume":"90","author":"Kass","year":"1995","journal-title":"J. Am. Stat. Assoc."},{"key":"ref_25","doi-asserted-by":"crossref","first-page":"1784","DOI":"10.1002\/psp4.13222","article-title":"A Fully Automatic Tool for Development of Population Pharmacokinetic Models","volume":"13","author":"Chen","year":"2024","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_26","doi-asserted-by":"crossref","first-page":"525","DOI":"10.1002\/psp4.12404","article-title":"NONMEM Tutorial Part I: Description of Commands and Options, with Simple Examples of Population Analysis","volume":"8","author":"Bauer","year":"2019","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_27","doi-asserted-by":"crossref","first-page":"161","DOI":"10.1002\/psp4.12742","article-title":"SAMBA: A Novel Method for Fast Automatic Model Building in Nonlinear Mixed-Effects Models","volume":"11","author":"Prague","year":"2022","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_28","doi-asserted-by":"crossref","first-page":"243","DOI":"10.1007\/s10928-021-09782-9","article-title":"Development of a Genetic Algorithm and NONMEM Workbench for Automating and Improving Population Pharmacokinetic\/Pharmacodynamic Model Selection","volume":"49","author":"Ismail","year":"2022","journal-title":"J. Pharmacokinet. Pharmacodyn."},{"key":"ref_29","doi-asserted-by":"crossref","first-page":"327","DOI":"10.1038\/s43856-025-01054-8","article-title":"A Machine Learning Approach to Population Pharmacokinetic Modelling Automation","volume":"5","author":"Richardson","year":"2025","journal-title":"Commun. Med."},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"558","DOI":"10.1208\/s12248-009-9133-0","article-title":"Importance of Shrinkage in Empirical Bayes Estimates for Diagnostics: Problems and Solutions","volume":"11","author":"Savic","year":"2009","journal-title":"AAPS J."},{"key":"ref_31","doi-asserted-by":"crossref","first-page":"17","DOI":"10.1038\/sj.clpt.6100241","article-title":"Diagnosing Model Diagnostics","volume":"82","author":"Karlsson","year":"2007","journal-title":"Clin. Pharmacol. Ther."},{"key":"ref_32","doi-asserted-by":"crossref","first-page":"187","DOI":"10.1007\/s10928-024-09921-y","article-title":"Evaluation of ChatGPT and Gemini Large Language Models for Pharmacometrics with NONMEM","volume":"51","author":"Shin","year":"2024","journal-title":"J. Pharmacokinet. Pharmacodyn."},{"key":"ref_33","doi-asserted-by":"crossref","first-page":"360","DOI":"10.1111\/bcp.15895","article-title":"ChatGPT in Pharmacometrics? Potential Opportunities and Limitations","volume":"90","author":"Cloesmeijer","year":"2024","journal-title":"Br. J. Clin. Pharmacol."},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"1965","DOI":"10.1002\/psp4.70125","article-title":"AI for NONMEM Coding in Pharmacometrics Research and Education: Shortcut or Pitfall?","volume":"14","author":"Zheng","year":"2025","journal-title":"CPT Pharmacomet. Syst. Pharmacol."},{"key":"ref_35","doi-asserted-by":"crossref","first-page":"34","DOI":"10.1007\/s10928-025-09982-7","article-title":"Leveraging Large Language Models in Pharmacometrics: Evaluation of NONMEM Output Interpretation and Simulation Capabilities","volume":"52","author":"Cha","year":"2025","journal-title":"J. Pharmacokinet. Pharmacodyn."}],"container-title":["Pharmaceutics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1999-4923\/18\/4\/501\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T04:24:21Z","timestamp":1777091061000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1999-4923\/18\/4\/501"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,18]]},"references-count":35,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2026,4]]}},"alternative-id":["pharmaceutics18040501"],"URL":"https:\/\/doi.org\/10.3390\/pharmaceutics18040501","relation":{},"ISSN":["1999-4923"],"issn-type":[{"value":"1999-4923","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,18]]}}}