Multiple machine learning models combined with virtual screening and molecular docking to identify selective human ALDH1A1 inhibitors

Abstract

Aldehyde dehydrogenases (ALDHs) are the enzymes of oxidoreductase family that are responsible for the aldehyde metabolism. The unbalanced expression of these enzymes may be associated with a variety of disease conditions including cancers. ALDH1A1 is one of the isoform of ALDHs majorly overexpressed in a variety of tumors and responsible for the anti-cancer drug resistance. This makes ALDH1A1 as a specific target to develop small molecule ALDH1A1 inhibitors for resistant cancer condition. Number of ALDH1A1 inhibitors have been developed and reported in the literature, but because of non-selectivity and inappropriate pharmacokinetic properties till now none of these have reached in the market for clinical use. Therefore, multiple machine learning models of different isoforms of ALDHs are integrated with in-silico techniques including virtual screening, docking, ADMET profiling, and MD simulation to identify selective ALDH1A1 inhibitors. Total ten selective ALDH1A1 inhibitors with diverse scaffolds and appropriate ADMET were identified that can be further developed as adjuvant therapy in cyclophosphamide and cisplatin resistance cancer.

Keywords: ADMET; ALDH1A1 inhibitors; Docking; Drug resistance; MD simulations; Multiple machine learning models; Virtual screening.