{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T06:04:08Z","timestamp":1790661848085,"version":"4.1.0"},"reference-count":62,"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. Pharmacol."],"abstract":"\n Accelerating the identification of viable drug candidates across therapeutic areas demands autonomous systems for molecular generation that enforce chemical correctness by structural invariants rather than model capability. We present MolecureAI, an autonomous AI agent architecture for closed-loop drug discovery with cost-ordered cascade evaluation that addresses four capability gaps unresolved across the field: autonomous design strategy selection, persistent cross-campaign structural memory, in-loop ADMET safety gating, and literature-scale evidence integration into the generative loop. A cost-ordered cascade concentrates full structure-based evaluation on a small pre-filtered survivor set, achieving a\n \n \n \n 2.5<\/mml:mn>\n \u00d7<\/mml:mo>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n reduction in candidates advanced relative to those generated. An in-loop metabolic gate enforces Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) safety by writing flagged reactive substructures to a per-target learned state, suppressing identified liabilities from all subsequent generation rounds. A persistent campaign memory influences generation trajectories by embedding prior structural success and failure patterns directly into the design context. We validate these architectural invariants across three structurally diverse benchmark targets (EGFR,\n \n \n \n \n \n KRAS<\/mml:mtext>\n <\/mml:mrow>\n \n G<\/mml:mtext>\n 12<\/mml:mn>\n C<\/mml:mtext>\n <\/mml:mrow>\n <\/mml:msup>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n , and BRD4) through a five-condition component ablation and a controlled memory ablation, and confirm generator agnosticism across four language-model backends\n via<\/jats:italic>\n a deterministic recalibration layer that overrides backend self-evaluation bias. Without explicit chemotype instruction, the pipeline computationally converged toward candidates bearing the 4-anilinoquinazoline scaffold characteristic of the approved inhibitors erlotinib and gefitinib, from structural feedback alone. Persistent campaign memory showed directional but non-significant trends toward lower inter-round score variance and higher late-stage advance rate (\n \n \n \n n<\/mml:mi>\n \u2248<\/mml:mo>\n 90<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n per condition; all\n \n \n \n p<\/mml:mi>\n ><\/mml:mo>\n 0.20<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:inline-formula>\n ). The component ablation reveals that optimizing for a single objective collapses performance on others; only full orchestration resolves this tension. Advancement rates above 79% across all backends confirm that the validated properties are generator-agnostic, enforced by architectural invariants rather than model capability. All reported outcomes are properties of the computational generation process; prospective biochemical validation of the advanced candidates remains the primary direction for future work.\n <\/jats:p>","DOI":"10.3389\/fphar.2026.1917774","type":"journal-article","created":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T05:50:00Z","timestamp":1790661000000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["MolecureAI: an agentic AI platform for autonomous drug discovery"],"prefix":"10.3389","volume":"17","author":[{"given":"Mohamed Amine","family":"Chadi","sequence":"first","affiliation":[{"name":"Laboratory of Computer Science and Smart Systems (LISI), Faculty of Sciences Semlalia, Cadi Ayyad University","place":["Marrakech, Morocco"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Asmaa","family":"Amzil","sequence":"additional","affiliation":[{"name":"Laboratory of Computer Science and Smart Systems (LISI), Faculty of Sciences Semlalia, Cadi Ayyad University","place":["Marrakech, Morocco"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hajar","family":"Ezzahoud","sequence":"additional","affiliation":[{"name":"Laboratory of Computer Science and Smart Systems (LISI), Faculty of Sciences Semlalia, Cadi Ayyad University","place":["Marrakech, Morocco"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hajar","family":"Mousannif","sequence":"additional","affiliation":[{"name":"Laboratory of Computer Science and Smart Systems (LISI), Faculty of Sciences Semlalia, Cadi Ayyad University","place":["Marrakech, Morocco"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,29]]},"reference":[{"key":"B1","volume-title":"Dziner: Rational Inverse Design of Materials with AI Agents","author":"Ansari","year":"2024"},{"key":"B2","volume-title":"Model Context Protocol: An Open Standard for Connecting AI Assistants to the Systems Where Data Lives. Tech. Rep","year":"2024"},{"key":"B3","doi-asserted-by":"crossref","first-page":"12004","DOI":"10.18653\/v1\/2025.emnlp-main.603","article-title":"LIDDIA: language-based intelligent drug discovery agent","volume-title":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","author":"Averly","year":"2025"},{"key":"B4","doi-asserted-by":"publisher","first-page":"2064","DOI":"10.1021\/acs.jcim.1c00600","article-title":"MolGPT: molecular generation using a transformer-decoder model","volume":"62","author":"Bagal","year":"2022","journal-title":"J. Chem. Inf. Model."},{"key":"B5","doi-asserted-by":"publisher","first-page":"2887","DOI":"10.1021\/jm9602928","article-title":"The properties of known drugs. 1. molecular frameworks","volume":"39","author":"Bemis","year":"1996","journal-title":"J. Med. Chem."},{"key":"B6","doi-asserted-by":"publisher","first-page":"235","DOI":"10.1093\/nar\/28.1.235","article-title":"The protein data bank","volume":"28","author":"Berman","year":"2000","journal-title":"Nucleic Acids Res."},{"key":"B7","doi-asserted-by":"publisher","first-page":"90","DOI":"10.1038\/nchem.1243","article-title":"Quantifying the chemical beauty of drugs","volume":"4","author":"Bickerton","year":"2012","journal-title":"Nat. Chem."},{"key":"B8","doi-asserted-by":"publisher","first-page":"525","DOI":"10.1038\/s42256-024-00832-8","article-title":"Augmenting large language models with chemistry tools","volume":"6","author":"Bran","year":"2024","journal-title":"Nat. Mach. Intell."},{"key":"B9","volume-title":"Mozi: Governed Autonomy for Drug Discovery LLM Agents","author":"Cao","year":"2026"},{"key":"B10","doi-asserted-by":"publisher","first-page":"65","DOI":"10.1186\/s13321-022-00643-2","article-title":"Conditional reduction of the loss value versus reinforcement learning for biassing a de-novo drug design generator","volume":"14","author":"Chadi","year":"2022","journal-title":"J. Cheminformatics"},{"key":"B11","volume-title":"Chemberta: Large-Scale self-supervised Pretraining for Molecular Property Prediction","author":"Chithrananda","year":"2020"},{"key":"B12","volume-title":"Clinicaltrials.Gov: A Resource Provided by the U.S","year":"2024"},{"key":"B13","doi-asserted-by":"publisher","first-page":"325","DOI":"10.1002\/prot.20552","article-title":"Comparing protein\u2013ligand docking programs is difficult","volume":"60","author":"Cole","year":"2005","journal-title":"Proteins Struct. Funct. Bioinforma."},{"key":"B14","doi-asserted-by":"publisher","first-page":"87","DOI":"10.1038\/nrd.2015.18","article-title":"The drug\u2013target residence time model: a 10-year retrospective","volume":"15","author":"Copeland","year":"2016","journal-title":"Nat. Rev. Drug Discov."},{"key":"B15","doi-asserted-by":"publisher","first-page":"6970","DOI":"10.1021\/jm050529c","article-title":"MetaSite: understanding metabolism in human cytochromes from the perspective of the chemist","volume":"48","author":"Cruciani","year":"2005","journal-title":"J. Med. Chem."},{"key":"B17","doi-asserted-by":"publisher","first-page":"5189","DOI":"10.1021\/acs.jcim.6c00535","article-title":"VeGA-RX and VeGA-SCX: controllable SMARTS-guided generative transformers for precision-driven de novo drug design","volume":"66","author":"Delre","year":"2026","journal-title":"J. Chem. Inf. Model."},{"key":"B18","doi-asserted-by":"publisher","first-page":"20","DOI":"10.1016\/j.jhealeco.2016.01.012","article-title":"Innovation in the pharmaceutical industry: new estimates of R&D costs","volume":"47","author":"DiMasi","year":"2016","journal-title":"J. Health Econ."},{"key":"B19","doi-asserted-by":"publisher","first-page":"2","DOI":"10.1186\/s13321-018-0324-5","article-title":"BioTransformer: a comprehensive computational tool for small molecule metabolism prediction and exploration","volume":"11","author":"Djoumbou-Feunang","year":"2019","journal-title":"J. Cheminformat."},{"key":"B20","doi-asserted-by":"publisher","first-page":"605","DOI":"10.1038\/nrd.2016.109","article-title":"Twenty years on: the impact of fragments on drug discovery","volume":"15","author":"Erlanson","year":"2016","journal-title":"Nat. Rev. Drug Discov."},{"key":"B21","doi-asserted-by":"publisher","first-page":"8","DOI":"10.1186\/1758-2946-1-8","article-title":"Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions","volume":"1","author":"Ertl","year":"2009","journal-title":"Mol. Inf."},{"key":"B22","doi-asserted-by":"publisher","first-page":"1230","DOI":"10.1021\/acsmedchemlett.8b00382","article-title":"Discovery of tetrahydropyridopyrimidines as irreversible covalent inhibitors of KRAS-G12C with in vivo activity","volume":"9","author":"Fell","year":"2018","journal-title":"ACS Med. Chem. Lett."},{"key":"B23","doi-asserted-by":"publisher","first-page":"39","DOI":"10.1007\/7355_2016_26","article-title":"Small molecule inhibitors of the epidermal growth factor receptor","volume":"1","author":"Finlay","year":"2017","journal-title":"Methods Mol. Biol."},{"key":"B46","doi-asserted-by":"publisher","first-page":"412","DOI":"10.1002\/minf.201700120","article-title":"Prediction of protein-compound binding energies from known activity data: docking-score-based method and its applications","volume":"35","author":"Fukunishi","year":"2018","journal-title":"Mol. Inform."},{"key":"B24","doi-asserted-by":"publisher","first-page":"841","DOI":"10.1016\/j.chembiol.2014.05.009","article-title":"Selective chemical modulation of gene transcription favors oligodendrocyte lineage progression","volume":"21","author":"Gacias","year":"2014","journal-title":"Chem. and Biol."},{"key":"B25","volume-title":"Txagent: An AI Agent for Therapeutic Reasoning Across a Universe of Tools","author":"Gao","year":"2025"},{"key":"B26","doi-asserted-by":"publisher","first-page":"109","DOI":"10.1609\/aaai.v38i1.27761","article-title":"Text-guided molecule generation with diffusion language model","volume":"38","author":"Gong","year":"2024","journal-title":"Proc. AAAI Conf. Artif. Intell."},{"key":"B27","volume-title":"3D Equivariant Diffusion for target-aware Molecule Generation and Affinity Prediction","author":"Guan","year":"2023"},{"key":"B28","doi-asserted-by":"publisher","first-page":"75","DOI":"10.1007\/s12551-013-0130-2","article-title":"Receptor\u2013ligand molecular docking","volume":"6","author":"Guedes","year":"2018","journal-title":"Biophys. Rev."},{"key":"B29","doi-asserted-by":"publisher","first-page":"D891","DOI":"10.1093\/nar\/gkad1049","article-title":"Ensembl 2024","volume":"52","author":"Harrison","year":"2024","journal-title":"Nucleic Acids Res."},{"key":"B30","doi-asserted-by":"publisher","first-page":"104605","DOI":"10.1016\/j.drudis.2026.104605","article-title":"Democratising real-world drug discovery through agentic AI","volume":"31","author":"He","year":"2025","journal-title":"Drug Discov. Today"},{"key":"B31","doi-asserted-by":"publisher","first-page":"e26724","DOI":"10.7554\/eLife.26724","article-title":"Systematic integration of biomedical knowledge prioritizes drugs for repurposing","volume":"6","author":"Himmelstein","year":"2017","journal-title":"eLife"},{"key":"B32","volume-title":"Biomni: A general-purpose Biomedical AI Agent","author":"Huang","year":"2025"},{"key":"B33","doi-asserted-by":"publisher","first-page":"583","DOI":"10.1038\/s41586-021-03819-2","article-title":"Highly accurate protein structure prediction with AlphaFold","volume":"596","author":"Jumper","year":"2021","journal-title":"Nature"},{"key":"B34","doi-asserted-by":"publisher","first-page":"D1373","DOI":"10.1093\/nar\/gkac956","article-title":"PubChem 2023 update","volume":"51","author":"Kim","year":"2023","journal-title":"Nucleic Acids Res."},{"key":"B35","doi-asserted-by":"publisher","first-page":"S142","DOI":"10.1016\/j.toxlet.2013.05.370","article-title":"Multi-parameter in vitro toxicity testing of crizotinib, sunitinib, erlotinib and other kinase inhibitors","volume":"219","author":"Krizovska","year":"2013","journal-title":"Toxicol. Lett."},{"key":"B36","volume-title":"Rdkit: Open-Source Cheminformatics","author":"Landrum","year":"2006"},{"key":"B38","volume-title":"Druggpt: A GPT-based Strategy for Designing Potential Ligands Targeting Specific Proteins","author":"Li","year":"2023"},{"key":"B39","doi-asserted-by":"crossref","DOI":"10.1145\/3770855.3819017","volume-title":"Molecular Lead Optimization via Agentic Tool Planning","author":"Li","year":"2026"},{"key":"B40","doi-asserted-by":"publisher","first-page":"3","DOI":"10.1016\/S0169-409X(96)00423-1","article-title":"Experimental and computational approaches to estimate solubility and permeability in drug discovery and development environments","volume":"23","author":"Lipinski","year":"1997","journal-title":"Adv. Drug Deliv. Rev."},{"key":"B41","doi-asserted-by":"publisher","first-page":"85","DOI":"10.1186\/s13321-021-00561-9","article-title":"DrugEx v2: de novo design of drug molecule by pareto-based multi-objective reinforcement learning in polypharmacology","volume":"13","author":"Liu","year":"2021","journal-title":"J. Cheminformat.s"},{"key":"B42","doi-asserted-by":"publisher","first-page":"20","DOI":"10.1186\/s13321-024-00812-5","article-title":"REINVENT 4: modern AI-driven generative molecular design","volume":"16","author":"Loeffler","year":"2024","journal-title":"J. Cheminformat."},{"key":"B16","doi-asserted-by":"publisher","first-page":"110417","DOI":"10.1016\/j.dib.2024.110417","article-title":"Dataset of drugs, their molecular scaffolds and medical indications with interactive visualization","volume":"15","author":"Malakhov","year":"2024","journal-title":"Nat. Commun."},{"key":"B43","doi-asserted-by":"publisher","first-page":"46563","DOI":"10.1021\/acsomega.4c08408","article-title":"CACTUS: chemistry agent connecting tool usage to science","volume":"9","author":"McNaughton","year":"2024","journal-title":"ACS Omega"},{"key":"B44","doi-asserted-by":"publisher","first-page":"D930","DOI":"10.1093\/nar\/gky1075","article-title":"ChEMBL: towards direct deposition of bioassay data","volume":"47","author":"Mendez","year":"2019","journal-title":"Nucleic Acids Res."},{"key":"B45","first-page":"80","article-title":"The rise of fragment-based drug discovery","volume":"4","author":"Murray","year":"2012","journal-title":"Nat. Chem."},{"key":"B47","doi-asserted-by":"publisher","first-page":"D1353","DOI":"10.1093\/nar\/gkac1046","article-title":"The next-generation open targets platform: reimagined, redesigned, rebuilt","volume":"51","author":"Ochoa","year":"2023","journal-title":"Nucleic Acids Res."},{"key":"B48","volume-title":"FROGENT: An end-to-end full-process Drug Design multi-agent System","author":"Pan","year":"2025"},{"key":"B49","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1038\/nrd.2018.168","article-title":"Drug repurposing: progress, challenges and recommendations","volume":"18","author":"Pushpakom","year":"2019","journal-title":"Nat. Rev. Drug Discov."},{"key":"B50","doi-asserted-by":"publisher","first-page":"656","DOI":"10.1038\/d41573-024-00068-6","article-title":"Efficiency, effectiveness and productivity in pharmaceutical R&D","volume":"23","author":"Roland","year":"2024","journal-title":"Nat. Rev. Drug Discov."},{"key":"B51","doi-asserted-by":"publisher","first-page":"D20","DOI":"10.1093\/nar\/gkab1112","article-title":"Database resources of the National Center for Biotechnology Information","volume":"50","author":"Sayers","year":"2022","journal-title":"Nuc. Acids Res."},{"key":"B52","doi-asserted-by":"publisher","first-page":"191","DOI":"10.1038\/nrd3681","article-title":"Diagnosing the decline in pharmaceutical R&D efficiency","volume":"11","author":"Scannell","year":"2012","journal-title":"Nat. Rev. Drug Discov."},{"key":"B53","doi-asserted-by":"publisher","first-page":"899","DOI":"10.1038\/s43588-024-00737-x","article-title":"Structure-based drug design with equivariant diffusion models","volume":"4","author":"Schneuing","year":"2024","journal-title":"Nat. Comput. Sci."},{"key":"B55","doi-asserted-by":"publisher","first-page":"3400","DOI":"10.1021\/acs.jcim.9b00376","article-title":"FAME 3: predicting the sites of metabolism in synthetic compounds and natural products for phase 1 and phase 2 metabolic enzymes","volume":"59","author":"\u0160\u00edcho","year":"2019","journal-title":"J. Chem. Inf. Model."},{"key":"B56","doi-asserted-by":"publisher","first-page":"btae416","DOI":"10.1093\/bioinformatics\/btae416","article-title":"ADMET-AI: a machine learning ADMET platform for evaluation of large-scale chemical libraries","volume":"40","author":"Swanson","year":"2024","journal-title":"Bioinformatics"},{"key":"B57","doi-asserted-by":"publisher","first-page":"455","DOI":"10.1002\/jcc.21334","article-title":"AutoDock Vina: improving the speed and accuracy of docking","volume":"31","author":"Trott","year":"2010","journal-title":"J. Comput. Chem."},{"key":"B58","doi-asserted-by":"publisher","first-page":"D523","DOI":"10.1093\/nar\/gkac1052","article-title":"UniProt: the universal protein knowledgebase in 2023","volume":"51","year":"2023","journal-title":"Nucleic Acids Res."},{"key":"B59","doi-asserted-by":"publisher","first-page":"2695","DOI":"10.1021\/ja512751q","article-title":"Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field","volume":"137","author":"Wang","year":"2015","journal-title":"J. Am. Chem. Soc."},{"key":"B60","volume-title":"Molmem: Memory-Augmented Agentic Reinforcement Learning for sample-efficient Molecular Optimization","author":"Wang","year":"2026"},{"key":"B61","doi-asserted-by":"publisher","DOI":"10.64898\/2025.12.17.694892","article-title":"BioChemAIgent: an AI-driven protein modeling and docking framework for structure-based","author":"Yousefi","year":"2025","journal-title":"Drug Discovery"},{"key":"B62","doi-asserted-by":"publisher","first-page":"5295","DOI":"10.1021\/acs.jcim.9b00727","article-title":"Deepscaffold: a comprehensive tool for scaffold-based de novo drug discovery using deep learning","volume":"59","author":"Zhang","year":"2019","journal-title":"J. Chem. Inf. Model."},{"key":"B63","doi-asserted-by":"publisher","DOI":"10.1101\/2025.06.03.657658","article-title":"OriGene: a self-evolving virtual disease biologist automating therapeutic target discovery","author":"Zhang","year":"2025"},{"key":"B37","doi-asserted-by":"publisher","first-page":"591","DOI":"10.1007\/s10822-022-00467-0","article-title":"Modeling receptor flexibility in the structure-based design of KRASG12C inhibitors","volume":"36","author":"Zhu","year":"2022","journal-title":"J. Computer-Aided Mol. Des."}],"container-title":["Frontiers in Pharmacology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fphar.2026.1917774\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,29]],"date-time":"2026-09-29T05:50:03Z","timestamp":1790661003000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fphar.2026.1917774\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,29]]},"references-count":62,"alternative-id":["10.3389\/fphar.2026.1917774"],"URL":"https:\/\/doi.org\/10.3389\/fphar.2026.1917774","relation":{},"ISSN":["1663-9812"],"issn-type":[{"value":"1663-9812","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,29]]},"article-number":"1917774"}}