Interpreting sliding windows in Ka/Ks analyses

Question

[tlorin]

Dear Biostars users,

I am running selection analysis on one gene on different species. I have performed whole sequence analyses using Ka/Ks calculator, and I was interested in finding the specific selected regions using sliding windows.

I have used the Matlab dN/dS tool for sliding windows (that is discussed here and for which tutorial is available here). I have picked a window size of 90 and get a figure for which region between positions 25 and 40 has a Ka/Ks ratio over 1. Cool! \o/ \o/

Here is the interpretation given on the Matlab page:

we observe several peaks over the threshold of 1. These regions appear to undergo positive selection that favors amino acid diversity, as it provides some fitness advantage.

And here is the problem: WHAT DOES IT MEAN? It surely means that there is a region under positive selection in my sequence. But WHERE?

1. Does it mean that THE WHOLE WINDOW between positions 25+90 and 40+90 is positively selected?
2. Does it mean that positive selection acts on 25-40 sites only?
3. How should I change my interpretation if I don't pick the "starting codon" visualization option, but the "middle codon" option?

You understand that picking option 1 or 2 would change completely the interpretation ;)

If anyone who has any experience with this software would like to share it it would be much appreciated!! :)

Answer 1

[karl.schmid]

Do not use sliding window analysis - they result in false positives: http://dx.plos.org/10.1371/journal.pone.0003746

Use site, branch-site models in PAML instead.