If evaluated by its longevity, the 55+ year old QSAR modeling can be properly called a senior citizen of the community of computational drug discovery approaches. Naturally, it is appropriate to both reminisce about the QSAR history and pontificate about its future. During its long history, QSAR has been both heavily criticized for problems such as model over-fitness and praised for critical contributions to the discovery of new drugs. One could observe that in the past decade QSAR modeling has become a stable and perhaps, even routine, technology as part of the computational medicinal chemistry toolkit. However, I will argue that in today’s world of big data, QSAR modeling is experiencing a reincarnation whereby many familiar features of the approach are finding much extended and expanded use. This transition is a result of a dramatic diversification of data types incorporated into traditional QSAR modeling schemas; growing use of sophisticated algorithms such as deep learning, and direct integration of QSAR with experimental drug discovery technologies. To exemplify these trends, I will present a novel consensus QSAR methodology called Multi-Descriptor Read Across (MuDRA), discuss recent studies on drug repurposing using transcriptomics data, and introduce a novel approach to biased chemical library generation using reinforcement learning. This talk is a tribute to Frank Brown’s life of a visionary scientist who always thought about the future using lessons learned from the past.

Reminiscing about the future of QSAR

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.