Title: Graph Neural Networks in Computational Biology: A Personal Perspective
Abstract:
Graph neural networks (GNNs) have in recent years become an invaluable tool for machine learning with data in the form of graphs or sets. GNNs now frequently form the back-bone for machine learning applications in drug discovery, modeling of small molecules, and more broader applications of computational biology.
I come from a graph representation learning background, but have done my PhD in a computational biology group -- this strongly exposed me to interactions with biologists, chemists and clinicians -- often leading to very interesting collaborations. Through them I've not only seen the transformative potential that graph representation learning could have in these areas, but also realised how low the barrier of entry currently is for interested deep learning practitioners. In some cases, the data is literally sitting there, waiting to be processed!
This talk will provide a brief introduction to GNNs and will cover my own journey through their applications in computational biology. Time permitting, this will span diverse areas such as antibody-antigen interactions, models of molecular interactions, small-molecule modelling, protein function prediction, algorithms for genome assembly, and patient outcome prediction in the intensive-care unit.
I will assume only a basic knowledge of machine learning with deep neural networks: no strong background knowledge in computational biology is required upfront.
24 авг 2024