Multiple Sequence Alignment quality comparison

Question

[Joe]

Does anyone know of any existing tools/scripts which can score the quality of MSAs? I have a couple of quite long alignments which I've done with Lastz, MAFFT, Clustalo and Kalign, but its basically impossible to check these by eye.

Crude options are valid too! Just after a metric or some kind of score/comparator I can use to choose the theoretically 'best' alignment.

[WouterDeCoster]

Perhaps some pairwise distance metric? Levenshtein/Hamming/... averaged,... just thinking out loud

[Joe]

Yeah that's the best idea I've had so far, and I've written a tool in the past to do a similar thing. The only other thing I can think of is the MSA with the lowest average entropy over all columns perhaps. Though I think gaps would need 'special treatment' as a column which was mostly gaps would have low entropy, despite being a fairly uninformative site.

[GenoMax]

See if this paper helps. Another one here.

[Joe]

I found SuiteMSA, but I don't think it was happy with the alignments I'd given it. It took ages to open them (via Xterm admittedly) and just ground to a halt. I might try it again locally... first paper looks good though thanks!

Answer 1

[Joseph Hughes]

I think the program MUMSA might be useful for this sort of thing. It has been designed to compare different alignments of the same sets of sequences and identifies regions of consistency between different alignments as well as identifying problematic regions (see ref).

Also in the PRANK paper, they use different methods for comparing alignments.