AI Guided Mutation Prediction and Vaccine Development Using Inverse Folding

Abstract

Vaccine and binder design relies on structural stability of the target protein. Being able to predict mutations or identify regions along the protein that do not mutate can help us better prepare for future pandemics. Since most mutations combined do not show extreme structural deviations (RMSD between original and mutant structure is under 3Å) we can use Inverse Folding techniques to identify regions that are stable or highly mutable based on positional entropies calculated from the resulting inverse folded sequences.