Coevolution Contact Map: How AI Reads Structure from Sequence

This is the statistical mechanism underneath AlphaFold: residues that touch in 3D tend to mutate together across a family of related sequences, so mining a multiple sequence alignment for mutual information between columns predicts which residue pairs are in contact — before any 3D coordinate exists. The simulator builds a synthetic protein with a known helix-turn-helix fold, generates an MSA carrying a coevolution signal (weakened by an adjustable mutation-noise rate), ranks residue pairs by mutual information, and uses only the top-K predicted contacts to drive a live spring-relaxation that folds an extended chain into 3D. Live readouts track contact precision against the known ground truth, the distance-RMSD of the folded model versus the true structure, and the mean MI strength of the contacts actually used — showing directly how MSA depth and noise determine whether the fold can be recovered at all.