Fig. 3

Expansion score (top) and SPFP (bottom) on large random backbones. In each subfigure, the top panel denotes biological datasets, and the bottom panel denotes simulated datasets (note the change in the y-axis range for expansion scores between the top and bottom panels). The horizontal dashed line indicates a perfect expansion score of 1. MAFFT-linsi--add failed to finish within 24 h for datasets except for 10AA and ROSE 1000M1–4 and was not run on Rec and Res (due to their very large size, we knew it would not complete within the allowed time), and MAFFT--add encountered out-of-memory issues on the Res dataset (failed runs, or runs that were not attempted, are marked with “X”). The expansion score is the length of the estimated alignment normalized by the length of the reference or true alignment; optimal is 1.0, and values above 1.0 indicate alignments that are too long (and so are under-aligned)