Effect of generating MergeAlign consensus MSAs with increasing numbers of constituent alignments. Individual MSAs were generated for a set of 1000 randomly selected benchmark MSAs using either MAFFT FFT-NS-2 (grey lines) or MAFFT L-INS-i (black lines). MSAs were added to the consensus MSA in the order in which the matrices that generated them are ranked ( Additional file 2). As a result, the mean F-score of constituent MSAs (marked with triangles) drops as more alignments are included. However, the mean F-score of MSAs generated by MergeAlign (marked with circles) is relatively stable. Furthermore, MergeAlign outperforms M-Coffee (dashed line) on the same set of benchmark MSAs.
Collingridge and Kelly BMC Bioinformatics 2012 13:117 doi:10.1186/1471-2105-13-117