"Engineering Drug Discovery Using Chemical Data Science and Immunology"
Current drug discovery has not matched the accelerating rate of technology development observed in many other walks of life, getting exponentially more expensive (costing billions of dollars) and less efficient (taking a decade or more of time). We have developed automated methods that will accelerate several areas of the drug design pipeline from synthesis, analytical characterization to desired bioactivity. Our approach employs deep learning compatible molecular representations (features) that reduce the time and cost of developing new molecules and processes while increasing their efficacy (desired property) because molecules (and related process) will be designed for specific properties rather than created using empirical knowledge. We will briefly discuss molecular representations and computational pipelines to develop a library of synthetically feasible bioactive molecules, and models for reactivity that will be used to predict and validate conditions for chemical reactions. We will also show a few examples of this combined model-based machine learning and experimental approach to select designs for function-specific, chemically diverse, potent and non-toxic lead molecules for solid tumors and suppressive immune cell types that are tested in vivo mouse models and in pet dogs’ clinical study. Finally, we expect to bring all these models together in a virtual reality-based drug design game that we have developed to crowdsource automated molecular discovery of molecules with desired properties. We expect that engineering drug design based on phenotype (property) of interest will save cost and time by accelerating different aspects of the drug discovery processes.
~Coffee/tea will be served prior to lecture~