Type 2 diabetes mellitus (T2DM) remains a major global health challenge, with many patients failing to achieve optimal glycaemic control despite the availability of established therapies. The enzyme 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1), which catalyzes the local regeneration of active cortisol, represents a promising therapeutic target for addressing the multifactorial nature of metabolic disorders.
This study integrates artificial intelligence and structural biophysics approaches to investigate the inhibitory potential of steroidal pregnanes against 11β-HSD1. A Random Forest-based quantitative structure–activity relationship (QSAR) model was first developed to identify key molecular determinants of bioactivity and prioritize candidate compounds. Subsequent molecular docking, 100 ns molecular dynamics simulations, and MM-GBSA binding free energy calculations were employed to elucidate the structural and energetic basis of ligand binding.
Two compounds, pregnane-3,20-diol disulphate (P1) and 20-piperidin-2-yl-5α-pregnan-3β,20-diol (P2), emerged as promising inhibitors, exhibiting high predicted bioactivity, favorable pharmacokinetic properties, and stable interactions within the NADPH-dependent catalytic cavity. Unlike the reference inhibitor carbenoxolone, which primarily forms peripheral polar contacts, these pregnane scaffolds exhibit deeper insertion into the hydrophobic pocket, and thereby enhancing shape complementarity and van der Waals packing.
Overall, this work provides mechanistic insights into 11β-HSD1 modulation and establishes a framework for the rational design of pregnane-based therapeutics. The study highlights the transformative role of AI-driven structural biophysics in accelerating drug discovery for metabolic diseases in Africa and beyond.
