Fig. 4

SPFN of EMMA, WITCH-ng-add, and MAFFT-linsi--add when adding sequences into large random backbone alignments. The top panel denotes biological datasets, and the bottom panel denotes simulated datasets. Error bars indicate standard errors. MAFFT-linsi--add only completed within the allowed 24 h on datasets with at most 1000 sequences, and we did not run it on Rec or Res due to their large numbers of sequences (failed runs, or runs that were not attempted, are marked with “X”). Comparisons with MAFFT--add can be found in Additional file 1: Fig. S7. SPFN results for each of the 10AA datasets are in Additional file 1: Fig. S12