Direct Coupling Analysis (DCA)

Direct Coupling Analysis is a statistical-inference method that reads a Multiple Sequence Alignment of a protein family and infers which residue pairs are directly coupled — disentangling direct from merely correlated (transitively linked) pairs by fitting a global maximum-entropy (Potts / pairwise Boltzmann) model to the alignment's low-order statistics. In protein biology it is the standard tool for predicting contacting/coevolving residues from sequence alone. The vault meets it as a benchmarking target: applied to an evolved allosteric-network ensemble treated as a synthetic alignment, its successes and its limits become measurable against a ground-truth model.

What this is

A pairwise, low-order inference over sequence statistics. Its power and its ceiling both follow from that: it recovers local couplings well, and it is structurally blind to dependencies that live only at higher order or across long range.

References

• claim-bravi-2020-applies-dca-to-evolved-allosteric-networks-as-synthetic-msa — DCA applied to a synthetic MSA of evolved allosteric networks.
• claim-dca-predicts-mutation-costs-but-poor-generative-model-of-allostery — accurate for point-mutation costs, poor as a generative model.
• claim-dca-underestimates-long-range-epistasis-in-allosteric-materials — captures short-range, underestimates long-range epistasis; traced to absent long-range correlations.
• claim-pdz-dca-couplings-track-short-range-epistasis-more-than-long-range — the PDZ-domain real-data validation.
• claim-dca-strongest-couplings-only-weakly-influenced-by-phylogenetic-bias — phylogenetic sampling bias mainly contaminates intermediate, not the strongest, couplings.