In silico prediction of endocrine activity

Endocrine-disrupting chemicals (EDCs) pose health risks; yet, conventional in vitro and in vivo testing remains slow, costly, and animal-intensive. Enhanced use of in silico approaches can contribute to earlier and increased detection of potential EDCs. This review evaluates advances in computational toxicology for early identification and prioritization of EDCs, with a focus on estrogen, androgen, thyroid, and steroidogenesis pathways. We discuss various implementations of ligand- and structure-based approaches, covering machine learning, deep learning, docking, molecular dynamics, and free energy methods, as well as their application in a regulatory setting. Moreover, we outline the future prospects necessary to advance the field and improve the in silico identification of potential EDCs.