Alberto Cazzaniga
(Area Science Park, Italy)

In the last years deep learning models inspired by Natural Language Processing have played a crucial role in advancing our understanding of proteins, ranging from accurate structure prediction to de novo design. In this seminar, I will discuss recent contributions of our Laboratory focused on interpreting and applying Protein Language Models: how does the study of the geometric properties of data representations help identifying remote relations among proteins? how can we leverage these models and co-evolution to reliably predict relevant physical quantities from sequences? how the geometry of the protein structure landscape help us navigate and organize structural predictions at scale, furthering our understanding of evolution?

Zoom registration link:
https://zoom.us/meeting/register/tJctfumpqTksE9aAvwRTItGnOmxUfczSmQGs