Abstract: Predicting the function of genes, proteins, and other molecular constituents of the cell is a difficult problem whose solution has potentially enormous benefit. The greatest successes in computational methods have come from AI/ML/Deep learning applied to sequence and structure, as well as non-AI methods such as molecular dynamics simulations. While these methods have enjoyed great success recently, the inference path from sequence to structure to function is long and full of pitfalls. Since the function of a molecule is essentially defined by its network of interactions--commonly called a network pathway--network comparison and analysis offers a more direct route to functional prediction. In this talk I will present the first successful functional predictions arising from purely topological cross-species network alignment. I will show how network alignment is capable of transferring knowledge of protein function between species not only without using sequence or structure, but even when no sequence similarity exists and no structural similarity is known. Thus, network alignment produces predictions that are orthogonal to and purely additive to existing prediction methods.
Bio: Wayne Hayes received his degrees in Astrophysics and Computer Science at the University of Toronto. His research spans the scales from molecular dynamics through cells to planets, solar systems, and galaxies. Before coming to UC Irvine, he worked with James Yorke, who shared the Japan Prize with Benoit Mandelbrot for co-developing Chaos Theory. He is currently an Associate Professor of Computer Science at UC Irvine. In his limited spare time, he flies airplanes, rides motorcycles, looks through telescopes, climbs mountains, and wrestles crocodiles.