Protein Sequence Coevolution & Mutual Information

Positions in a protein that interact structurally tend to coevolve — mutations at one site are compensated by mutations at its partner. Mutual information MI(i,j) = Σ p(a,b) log[p(a,b)/p(a)p(b)] reveals these couplings from a multiple sequence alignment.

Simulation

Seq length L: 30 Sequences M: 200 Coupling strength: 0.60 Noise: 0.20

MI measures statistical dependency between positions. High MI pairs (red) are candidate structural contacts. APC correction removes phylogenetic background.

Protein alphabet: 20 amino acids. Colors map to MI strength.