Why equivariance is so popular in Life and Material Science?

By Mario Geiger Senior Research Scientist, NVIDIA

Equivariant models like NequIP, MACE, Allegro, DiffDock, Equiformer, and EquiFold have significantly advanced Life and Material Sciences. These models effectively predict and understand complex systems by using symmetries in the data, addressing problems such as molecular dynamics, protein folding, and drug docking. This talk will review the latest progress in equivariant modelling, discussing how NequIP and MACE enhance material property predictions, Allegro and DiffDock improve molecular docking and dynamics, and Equiformer and EquiFold advance structural biolog