Coevolutionary analyses require phylogenetically deep alignments and better null models to accurately detect inter-protein contacts within and between species

Table 1 List of methods benchmarked

Coevolution methods benchmarked fall into three categories. Information-based methods: MI: mutual information [71], VI: variation of information [65], MI_j: MI divided by alignment column-pair entropy, MI_Hmin: MI divided by minimum column entropy [8], MI_w: MI with adjusted amino acid probabilities. Direct methods: DI: direct information—MI with re-estimated joint probabilities [9], DI₂₅₆, DI₃₂: DI using Hopfield-Potts for dimensional reduction (256 and 32 patterns respectively) [68], DI_plm: Frobenius norm of coupling matrices in 21-state Potts model using pseudolikelihood maximization [72], PSICOV: sparse inverse covariance estimation [14]. Phylogenetic methods: CoMap P-values for four analyses CMP_cor: substitution correlation analysis [10], CMP_pol for polarity compensation, CMP_chg for charge compensation, CMP_vol for volume compensation [2]

ISSN: 1471-2105