COCOA-Tree: Phylogenetic visualization and comparative analysis of coevolving residues

The evolutionary co-occurrence of amino acid changes between protein residues underlies key structural and functional properties of protein families. Building on these coevolutionary patterns, methods have been developed to identify groups of residues associated with enzyme functionalities, such as Statistical Coupling Analysis (SCA) or Specificity-Determining Position (SDP) methods. These methods and their variations differ in the metrics used to quantify coevolution, residues weighting schemes, and corrections introduced to mitigate noise and phylogenetic biases. Yet, systematic comparisons across methods are rarely performed, and the evolutionary origins of the coevolution patterns highlighted by each approach are seldom addressed, limiting our ability to disentangle functional from phylogenetic contributions. To address these issues, we introduce COCOA-Tree, a Python library for SCA-like dimensionality-reduction analyses. COCOA-Tree supports custom metrics and enables visualization of coevolutionary patterns on phylogenetic trees. We also provide guidance to map results onto 3D structures in PyMOL. Using COCOA-Tree, we reanalyze published datasets and uncover previously unnoticed evolutionary properties of groups of coevolving residues detected by SCA, known as sectors. In particular, in the well-studied S1A serine protease family, we show that two of the three known sectors exhibit qualitatively distinct levels of sequence conservation depending on the enzymatic functions and on the phylogenetic clades to which the proteins belong. We further show that different coevolution metrics often identify qualitatively distinct groups of coevolving residues, although they yield consistent results for mildly conserved residues. Finally, as an example of the versatility of COCOA-Tree, we provide an example of visualizing pairs of residues detected by Direct Coupling Analysis (DCA) methods on a phylogenetic tree, highlighting a rich diversity of co-evolutionary patterns. Overall, we expect COCOA-Tree to help identify residues that control protein function and thereby improve our capacity for functional engineering and our understanding of the principles governing protein evolution. COCOA-Tree website: https://tree-timc.github.io/cocoatree