Journal Article Ahead of print
PubMed: 41565685 DOI: 10.1038/s41598-025-23764-8
Ligand-based machine learning models to classify active compounds for prostaglandin EP2 receptor.
Sci Rep
Authors: Paul Dupuyds, Anita Rayar, Vaitea Opuu, Camille Lescanne, Shuhui Wang, Jean-Philippe Annereau, Philippe Nghe, Luis BriseƱo-Roa
Affiliations
- Medetia Pharmaceuticals, 24 Boulevard du Montparnasse, 75015, Paris, France. Laboratory of Biophysics and Evolution, UMR CNRS-ESPCI 8231 Chimie Biologie Innovation, PSL Research University, ESPCI Paris, Paris, France.
- Medetia Pharmaceuticals, 24 Boulevard du Montparnasse, 75015, Paris, France.
- Laboratory of Biophysics and Evolution, UMR CNRS-ESPCI 8231 Chimie Biologie Innovation, PSL Research University, ESPCI Paris, Paris, France.
- Medetia Pharmaceuticals, 24 Boulevard du Montparnasse, 75015, Paris, France.
Abstract
Prostaglandin receptors are pharmacologically validated targets with implications in several medical indications, including glaucoma, cardiac cyanotic disease, pulmonary hypertension, oncology, and various rare diseases. In this study, we developed a ligand-based machine learning (ML) model to classify chemical compounds as either active or inactive against the prostaglandin receptor EP2. From an initial set of 1,826 descriptors, 20 were selected to train random forest algorithms, yielding an area under the curve score (AUC) ofĀ >Ā 0.8 for compound classification in the test set. Our resulting ML classifier showed an overall accuracy of 88.9% towards newly experimentally tested EP2 ligands. This adaptable and tractable workflow can be extended to other EP receptors and possibly other similar targets.
MeSH
Prostaglandins, Machine Learning, Natural Science Disciplines, Learning, Receptors, Prostaglandin, Ligands