Predicting the Effects of Mutations With Deep Generative Models
Adam Riesselman, Harvard University
Modern genome sequencing and synthesis can acquire and generate tremendous molecular diversity in a day, but our ability to navigate and interpret the exponentially large space of potential biological sequences remains limited. Central to this challenge is the lack of <em>a priori</em> knowledge about epistasis, i.e. non-additive interactions between positions in a molecule or genome. I will describe how generative models fit to evolutionary sequences can be used to help explain these factors. I will then discuss two classes of generative models, discrete undirected graphical models and neural-network powered latent variable models, that can reveal the three-dimensional structures and mutational landscapes of proteins and RNA solely from evolutionary information.
Abstract Author(s): Adam Riesselman, John Ingraham, Debora Marks