Schedule 25 September 2020 (times in British Summer Time BST):

17.30-18.30 BST

Guadalupe Gonzalez-Pigorini (Imperial College London): Applications of graph neural networks in computational biology

Abstract: In many instances, the complex machinery underlying cellular processes in the human body can be represented as a network. Metabolic networks, signalling networks, neuronal networks and protein-protein interaction networks are examples of this. This means that a great number of data sources used in biomedical research have an underlying network structure that is often not fully exploited by computational methods. Even though advances in machine learning have made their way into biomedical research, most of the approaches used analyze biological data disregarding its underlying network structure or using simple or hand-engineered algorithms to account for the interactions between features. Graph neural networks (GNNs) provide an excellent framework for analyzing biomedical data together with its underlying graph structure given that they accept a graph and its associated signal as input features. In this talk, I will present successful works on application of GNNs in computational biology with an emphasis on problems using genomic data. I will also present a work in progress to predict drug combinations against COVID-19.