Exploring sodium glucose cotransporter (SGLT2) inhibitors with machine learning approach: A novel hope in anti-diabetes drug discovery.

Conventional anti-diabetes agents exhibit some undesirable side effects. Recently, lactic acidosis and/or bladder cancer were also reported with the use of these agents. Hence, there is an urgent need for alternative anti-diabetes in order to reduce/avoid the unwanted effects. In this scenario sodium glucose cotransporter 2 (SGLT2) inhibitors has already been established as an important class of anti-diabetic drug. The search for new generation SGLT2 inhibitors with high affinity is still an ongoing process. Here, we aim to develop computational models to predict the SGLT2 inhibitory activity of small molecules based on chemical structures. This work provides in-silico analysis to propose possible fragment/fingerprint identification recommended for SGLT2 inhibitors. Up-to-our knowledge, this study is an initiative to propose fingerprints responsible for SGLT2 inhibition. Furthermore, we used nine different algorithms to build machine learning (ML) models that could be used to prioritize compounds as SGLT2 inhibitors from large libraries. The best performing ML models were applied to virtually screen a large collection of FDA approved drugs. The best predicted compounds have been recommended to be biologically investigated in future in order to identify next generation SGLT2 inhibitors with different chemical structure.