Computational Analysis and Classification of Vitamin D Receptor Ligands

Abstract

The vitamin D receptor (VDR) is a ligand-dependent nuclear transcription factor that plays a key role in regulating calcium homeostasis, immune processes, and cellular differentiation, and is a recognized target in drug development. Despite the large number of described VDR ligands, so far there has been no standardized collection with computational annotation suitable for system analysis. In this work, a curated dataset was formed and characterized, including 962 VDR ligands divided into four structural categories: analogues, metabolites, hybrid compounds and non-secosteroidal compounds, digitized based on the Maestro and Seoane catalog. A comprehensive chemoinformatic approach was used for the analysis, including the calculation of physico-chemical descriptors, the index of similarity to natural compounds, as well as four types of molecular fingerprints. Next, similarity analysis, clustering using the Butina algorithm, and visualization of the chemical space using t-SNE and UMAP methods were performed. The results showed that the studied set occupies a relatively compact and predominantly lipophilic chemical space, largely determined by the common secosteroid framework. At the same time, non-secosteroidal ligands are clearly separated on the projections of the chemical space, which confirms their belonging to a structurally distinct chemotype. The formed data set and the proposed multi-level analytical approach. In addition, the analyzed data set based on various types of molecular fingerprints forms a reproducible basis for further modeling the activity of VDR ligands, studying the structure–activity relationship, and rationally designing new analogues.