Hi Sergei,

Before using directionalrel.bf on my own data, I thought I would try to reproduce some of the results you published in "A Maximum Likelihood Method for Detecting Directional Evolution in Protein Sequences and Its Application to Influenza A Virus", and I seem to be running in to problems.

Taking the h3.seq.prot dataset, and the h3.seq.prot.rooted tree, I run directionalrel.bf with the Flu H5N1 substitution model, and specify the first sequence in the file to be the root. I've verified that the first sequence is indeed the root of the tree, and from its accession number AF348176 I can see that it is A/Hong Kong/1/68, which makes sense.

The results I get (attached) look reasonable, and they contain some locations identified in the published paper, such as 245, from which I can verify the alignment, but there are other locations identified which are not in the paper, and published locations which are not in the result set.

Is it possible that the method in HyPhy has been refined since publication, so that some differences in results might be expected, or have I done something wrong in conducting the analysis?

Many thanks for your help

William

Hi Sergei,

Many thanks for your reply. I repeated the analysis with the REV substitution model, which, not surprisingly, took quite a bit longer to run.

The highlighted sites are a near subset of those I obtained with the H5N1 substitution model, and still quite different from those published in the study. It is also striking that 11 out of the 25 sites identified in my two results sets have isoleucine as the preferred residue. I've attached a small comparison of results which I hope is understandable.

For unrelated reasons I've decided that DEPS is probably not the right tool for my work, but I thought I would pass these results on in case you wished to confirm that the analysis was still functioning as expected.

Regards,

William