Unravelling the multifaceted actions of neurosteroids: Machine learning and in vitro screening for novel target discovery

Abstract

Background and purpose: Neurosteroids (NS) modulate neuronal function and are promising therapeutic agents for neuropsychiatric disorders. NS analogues are approved for treating postpartum depression and are of interest in other disorders. Gamma-aminobutyric acid type A (GABAA) receptors are well characterised targets of natural NS, but other biological pathways are likely relevant to therapeutic mechanisms and/or to off-target effects. Experimental Approach: We performed hypothesis-generating in silico analyses using machine learning and broad in vitro biological screens to assess the range of actions of NS analogues of varying structural attributes. We employed machine learning for molecular similarity analysis and network pharmacology to elucidate likely targets. Key Results: This analysis confirmed likely targets beyond GABAA receptors. We then functionally screened 19 distinct NS structures across 78 targets representing interconnected signalling pathways, complemented with a limited screen of kinase activation. Results revealed the unanticipated modulation of targets by natural and synthetic NS analogues. Many compounds initiated androgen receptor (AR) translocation, with little or no enantioselectivity. Modulation of multiple G-protein receptors also was unexpected. Conclusions and Implications: NS compounds are ascendant treatments in neuropsychiatry, but their full spectrum of actions remains unclear. This virtual and biological screening discovery approach opens new vistas for exploring the mechanism of NS analogues. The multifaceted approach provides an unbiased, holistic exploration of the potential effects of NS compounds across various molecular targets and provides a platform for future validation studies to aid drug discovery.