QSAR for Vitamin D Receptor Ligands Prediction Through Machine Learning

Abstract

Osteoporosis is a highly prevalent bone disease characterized by reduced bone mineral density and an increased risk of fractures, particularly among postmenopausal women. Globally, approximately one in three women over 50 years of age is affected. In Colombia, women account for 92% of reported osteoporosis cases, and vitamin D deficiency is a common aggravating factor. Vitamin D regulates bone metabolism through the Vitamin D Receptor (VDR), making this pathway an important therapeutic target. However, prolonged use of vitamin D analogs may cause adverse effects such as hypercalcemia, limiting their clinical application. To address this challenge, this study proposes a computational framework based on Quantitative Structure-Activity Relationship (QSAR) for the discovery of novel VDR agonists. The approach integrates Machine Learning models trained on molecular descriptors, structural fingerprints, and experimental bioactivity data to identify and prioritize promising compounds. Candidate molecules are subsequently evaluated through molecular docking to estimate their binding affinity to VDR and through ADMET analyses to assess their pharmacological safety profiles. By combining predictive modeling with structure-based validation, this strategy aims to accelerate the identification of potential therapeutic candidates for osteoporosis while reducing the cost and time associated with experimental screening.