FIG 8.

ROC curves demonstrating that the statistical fitness is indicative of variant activity. The experimentally observed outcome (retention or ablation of catalytic activity) for each qualifying variant was evaluated against the variant’s statistical fitness, calculated using one of the six sets of protein sequences in the starting MSA (Table 1). “(L)” denotes that the lax cutoff threshold for outlier detection was used, and “(S)” denotes that the stringent cutoff threshold was used. The AUC consistently improved when restrictions placed on the protein sequences used in the MSA, in terms of both allotted sequence length and taxonomy, were relaxed. The best predictive method employed the least constrained MSA containing the most sequence information.