Deep neural network modeling based virtual screening and prediction of potential inhibitors for renin protein

Abstract

Renin enzyme plays an essential role in the Renin-Angiotensin System (RAS), and it is involved in the pathogenesis of hypertension and several other cardiovascular diseases (CVDs). Inhibition of renin is an effective way to intervene with the pathogenesis of these diseases. Docking-based virtual screening, 3D-Quantitative Structure-Activity Relationship (3D-QSAR), and structure-based drug design are the most frequently used strategies towards discovering novel inhibitors targeting renin. In this study, we have developed a 2D fingerprint-based Deep Neural Network (DNN) classifier for virtual screening and a DNN-QSAR model for biological activity prediction. The resulting hits from the DNN-QSAR model were then subjected to the molecular docking to identify further top hits. Molecular Dynamics (MD) simulation was conducted to get a better insight into the binding mode of identified hits. We have discovered six compounds from the Maybridge chemical database with the predicted IC₅₀ values ranging from 24.2 nM to 83.6 nM. To the best of our knowledge, this is the first study that used a cascaded DNN model to identify potential lead compounds for the inhibition of renin target. Through the results presented in this study, we provide evidence of the DNN method being a useful approach to identify new chemical entities/novel lead compounds that may overcome the limitation of existing conventional strategies used in drug discovery research.

Communicated by Ramaswamy H. Sarma

Keywords:

• Activity prediction
• deep neural network
• MD simulation
• molecular docking
• QSAR

Acknowledgements

SB acknowledges University Grant Commission for research fellowship grant. This manuscript has CSIR-CDRI manuscript number 10159.

Disclosure statement

None was reported by the authors.