Harnessing Deep Learning for Drug Discovery: Challenges and Opportunities

Speaker: Francesca Grisoni (TU/Eindhoven)

Abstract:

Deep learning has had an incredible impact in various fields of science and technology, such as protein structure prediction and organic reaction planning. However, machine learning techniques are known to be particularly effective when large-scale datasets are available. Drug discovery, on the other hand, is usually a low-data endeavour, e.g., when it comes to hit finding and de novo molecule design – which limits the potential of ‘out-of-the-box’ machine learning approaches. This talk will reflect on lessons learned in applying machine and deep learning to low-data drug discovery, for bioactivity prediction and de novo drug design. Moreover, it will highlight some successful learning strategies to alleviate these challenges and accelerate the discovery of bioactive molecules.e.g., when it comes to hit finding and de novo molecule design – which limits the potential of ‘out-of-the-box’ machine learning approaches. This talk will reflect on lessons learned in applying machine and deep learning to low-data drug discovery, for bioactivity prediction and de novo drug design. Moreover, it will highlight some successful learning strategies to alleviate these challenges and accelerate the discovery of bioactive molecules.

This seminar will take place in room C4.174