Machine Learning-Guided QSAR Modeling and Virtual Screening of Medicinal Plant-Derived Phytochemicals for the Identification of Potential hDHFR Inhibitors
- 1Department of Pharmacy
- 2University of Science and Technology Chittagong
Abstract
Human dihydrofolate reductase (hDHFR) is a well-established anticancer drug target due to its essential role in folate metabolism and DNA synthesis. This study employed integrated machine based structural-activity relationship, virtual screening, and molecular docking to identify potential hDHFR inhibitors from medicinal plant-derived phytochemicals. A curated dataset of 1,368 compounds was obtained from ChEMBL and presented using Morgan fingerprints and RDKit molecular descriptors. Four regression model algorithms, including RF, XGBoost, LightGBM, and SVR were developed and optimized. Among them, SVR achieved the best score with a test R2 of 0.669 and a Spearman coefficient of 0.789. The SVR model was then used to screen the drug like phytocompounds collected from medicinal plants. The top-ranked compounds were subsequently evaluated through molecular docking against hDHFR (PDB ID: 1U72). Several phytochemicals exhibited stronger binding affinities than the reference inhibitor methotrexate. Among them Withaferin A showed the strongest docking score (-11.68 kcal/mol). These findings highlight the potential of combining machine learning and structure based virtual screening to identify promising hDHFR-targeted lead compounds for further validation.
Related
- Discovery of potential ALK inhibitors from Glycosmis phytochemicals via integrated machine learning and umbrella sampling approaches
- Investigating the anticancer potential of phytochemicals from the mangrove Rhizophora mucronata as inhibitors of lung cancer-associated proteins: Insights from multiscale computational and in silico analyses
- Integrated machine learning, pharmacophore screening, and structure-based computational approaches for the identification of novel JAK3 inhibitors
- Structure-based virtual screening identifies naturally derived histidine kinase VbrK inhibitors as potential countermeasures against β-lactam antimicrobial resistance
- Integrated virtual screening and in vitro evaluation identify potential c-Myc/max complex inhibitors for colorectal cancer
BibTeX
@misc{sarker2026machine,
title = {{Machine Learning-Guided QSAR Modeling and Virtual Screening of Medicinal Plant-Derived Phytochemicals for the Identification of Potential hDHFR Inhibitors}},
author = {Shawon Sarker},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15010020/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010020/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Machine Learning-Guided QSAR Modeling and Virtual Screening of Medicinal Plant-Derived Phytochemicals for the Identification of Potential hDHFR Inhibitors
AU - Shawon Sarker
PY - 2026
DA - 2026/10/07
PB - ChemRxiv
DO - 10.26434/chemrxiv.15010020/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010020/v1
AB - Human dihydrofolate reductase (hDHFR) is a well-established anticancer drug target due to its essential role in folate metabolism and DNA synthesis. This study employed integrated machine based structural-activity relationship, virtual screening, and molecular docking to identify potential hDHFR inhibitors from medicinal plant-derived phytochemicals. A curated dataset of 1,368 compounds was obtained from ChEMBL and presented using Morgan fingerprints and RDKit molecular descriptors. Four regression model algorithms, including RF, XGBoost, LightGBM, and SVR were developed and optimized. Among them, SVR achieved the best score with a test R2 of 0.669 and a Spearman coefficient of 0.789. The SVR model was then used to screen the drug like phytocompounds collected from medicinal plants. The top-ranked compounds were subsequently evaluated through molecular docking against hDHFR (PDB ID: 1U72). Several phytochemicals exhibited stronger binding affinities than the reference inhibitor methotrexate. Among them Withaferin A showed the strongest docking score (-11.68 kcal/mol). These findings highlight the potential of combining machine learning and structure based virtual screening to identify promising hDHFR-targeted lead compounds for further validation.
ER -