In Direct Coupling Analysis, the strongest inferred couplings are only weakly influenced by phylogenetic bias — the confound concentrates in intermediate-strength couplings instead

Direct Coupling Analysis (DCA) infers residue-residue couplings from a multiple sequence alignment, and a standing critique of the method is that shared ancestry — phylogenetic correlation among sequences, rather than genuine coevolution — can inflate or distort inferred couplings, motivating pseudocount regularization and phylogenetic reweighting as standard corrections. Rodriguez Horta & Weigt (2021, PLOS Comp Biol) tested where that confound actually lands rather than assuming it applies evenly. Their finding runs against the naive expectation that the biggest signals are the most exposed to bias: "the largest coevolutionary couplings... are only weakly influenced by phylogeny." Spurious phylogenetic correlation instead concentrates in the intermediate-strength, false-positive-adjacent couplings — not in the strongest, structurally meaningful ones.

This bears directly on the vault's entity-direct-coupling-analysis cluster, specifically on how to read claim-dca-underestimates-long-range-epistasis-in-allosteric-materials and claim-pdz-dca-couplings-track-short-range-epistasis-more-than-long-range's long-range weakness. If phylogenetic bias mainly contaminated the strongest couplings, it would offer a tidy artifact explanation for why DCA misses strong long-range epistasis: correct for the bias and the missing signal reappears. Rodriguez Horta & Weigt's result argues the opposite — the strongest couplings, long-range or not, are largely phylogeny-clean already, which weighs against reading DCA's long-range blind spot as a correctable sampling artifact and toward reading it as a structural limit of pairwise, low-order inference itself.