Deep SAR matrix: SAR matrix expansion for advanced analog design using deep learning architectures

Abstract

Aim: Enhancing the structure–activity relationship matrix (SARM) methodology through integration of deep learning and expansion of chemical space coverage. Background: Analog design is of critical importance for medicinal chemistry. The SARM approach, which combines systematic structural organization of compound series with analog design, is put into scientific context. Methodology: The new DeepSARM concept is introduced. The architecture of SARM-integrated deep generative models is detailed and the workflow for advanced analog design and matrix expansion described. Exemplary application: The DeepSARM approach is applied to design analogs of kinase inhibitors taking kinome-wide chemical space into account. Future perspective: Practical applications of DeepSARM will be a major focal point. Different applications are discussed. New computational features will be added to prioritize virtual candidate compounds.