Figure 5

Distribution of F-measures obtained with each method for all alphabets for detecting individual coevolving positions. Modified box plots illustrate the distribution of F-measures obtained with a given method over all alphabets. Values are F-measure area under the curve (AUC) scores for (A) tetrapod myoglobin, (B) randomized tetrapod myoglobin, (C) chordate myosin rod, and (D) randomized chordate myosin rod. The computation of AUC values is described in Methods. Red lines indicate the median AUC, and the top and bottom of the boxes indicate the upper and lower quartile values, respectively. Whiskers represent the largest and smallest AUC values within 1.5 × IQR (inter-quartile range), and pluses represent outliers, or points outside of 1.5 × IQR. Methods with more condensed distributions are those that appear more robust to alphabet choice. AS: Ancestral States; L07: LnLCorr07; L99: LnLCorr99; MI: Mutual Information; NMI: Normalized Mutual Information; RMI: Resampled Mutual Information; Sn: Statistical Coupling Analysis, cutoff = n/10; MIP, Corrected Mutual Information; Gn: Generalized Continuous-Time Markov Process Coevolutionary Algorithm, ε = n/10; CM: CoMap.