In silico drug discovery framework leveraging biomedical foundation models and evaluating generative models—a use case on 3CLpro
Journal of Computer-Aided Molecular Design · 13.11.2026 · 40 (1)
DOI: 10.1007/s10822-026-00973-5 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Computer-Aided Molecular Design, online erschienen am 13.11.2026. Verfasst von Leili Zhang und 17 weiteren Autor:innen. Beteiligte Einrichtung: IBM Research, Yorktown Heights, NY, 10598, USA. zhangle@us.ibm.com. Hauptthemen laut MEDLINE: Drug Discovery. Die Arbeit nennt 44 Literaturangaben.
Abstract (Originalsprache)
AI models such as foundation models and generative models have shown promise for aiding drug discovery alongside traditional cheminformatics and bioinformatics tools and physics-based approaches. In this study, we design a workflow that integrates these AI methods to generate molecules alongside established informatic and predictive filtering approaches to streamline downselection of molecules from the massive and noisy result sets generated by AI. This workflow evaluates generative methods based on validity, reliability, diversity, similarity, hallucination and campaign needs. Analyses are made on the original proprietary dataset to determine molecular property thresholds for screening criteria. Overall, 249 molecules are selected from ~ 5 million generated molecules. 18 molecules are then validated with free energy perturbation simulations. 5 out of 5 retain on target inhibition, with IC50 values below 1 μM. 1 of the 5 experimentally validated molecules displayed greater inhibition than the corresponding lead molecule. Predictive foundation model MMELON yields Pearson correlation R = 0.72 in blind test, showing potential for future drug discovery campaigns.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- IBM Research, Yorktown Heights, NY, 10598, USA. zhangle@us.ibm.com.
- IBM Research, Yorktown Heights, NY, 10598, USA.
- Center for Therapeutics Discovery, Cleveland Clinic Research, Cleveland, OH, 44106, USA.
- Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Case Western Reserve University, Cleveland, OH, 44106, USA.
- Center for Therapeutics Discovery, Cleveland Clinic Research, Cleveland, OH, 44106, USA. Stauffs2@ccf.org.
- Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Case Western Reserve University, Cleveland, OH, 44106, USA. Stauffs2@ccf.org.
Zitieren
APA
Zhang, L., Kang, S. G., Huynh, T., Goins, C. M., Elsharkawy, L., Morrone, J. A., Burch, M., Premnauth, G., Rohde, J., Reilly, B., Fullenkamp, C. R., Rico, A., Ji, J., Coker, J. A., Weber, J. K., Wang, N., Stauffer, S. R., & Cornell, W. (2026). In silico drug discovery framework leveraging biomedical foundation models and evaluating generative models—a use case on 3CLpro. Journal of Computer-Aided Molecular Design, 40(1). https://doi.org/10.1007/s10822-026-00973-5
Vancouver
Zhang L, Kang SG, Huynh T, Goins CM, Elsharkawy L, Morrone JA, et al. In silico drug discovery framework leveraging biomedical foundation models and evaluating generative models—a use case on 3CLpro. Journal of Computer-Aided Molecular Design. 2026;40(1). doi:10.1007/s10822-026-00973-5
BibTeX
@article{Zhang2026,
author = {Zhang, Leili and Kang, Seung-Gu and Huynh, Tien and Goins, Christopher M and Elsharkawy, Lina and Morrone, Joseph A and Burch, Myson and Premnauth, Gurdat and Rohde, Joseph and Reilly, Brenna and Fullenkamp, Christopher R and Rico, Annabella and Ji, Jenny and Coker, Jesse A and Weber, Jeffrey K and Wang, Nancy and Stauffer, Shaun R and Cornell, Wendy},
title = {In silico drug discovery framework leveraging biomedical foundation models and evaluating generative models—a use case on 3CLpro},
journal = {Journal of Computer-Aided Molecular Design},
year = {2026},
volume = {40},
number = {1},
doi = {10.1007/s10822-026-00973-5}
}
EUVIMED