Hi,

I'm currently using datamonkey for my thesis on research on a HIV-1 cluster of 70 patients.
On one dataset, the SLAC, FEL and REL analysis gave following results:
     SLAC      REL      FEL      All 3
POS      8      64      20      8
NEG      36      36      44      22

My first question: for this moment I'm taking only the sites under selection that are present in all tree the analysis. Is this a misinterpretation? Do I then reject to many possible sites under selection?

Second question:
For another dataset, I get similar results on the SLAC and FEL analysis, but the REL gives very strange results. The sites that are under positive selection in SLAC and FEL came back as negative selected sites in the REL analysis and the same for the negative selected sites.
Can anyone explain this discordance ?

Many thanks
Kenny

Hi Kenny,

To find positive selection, I would recommend using a new method MEME, available on Datamonkey. REL is not particularly reliable when calling individual sites under selection, especially for smaller data sets, and we are in the process of replacing it with a more robust alternative in the next few months. For negative selection, please use FEL.

Generally SLAC < FEL < REL in terms of power to detect sites (e.g. in your row 1 this is a clear pattern), so intersecting the set of all sites will usually reduce to whatever SLAC found (which will be the most conservative).

Sergei

Hi Sergei,

Thanks for the quick reply.
Regarding my second question about the discordance between the FEL and REL results (positve in one method, negative in the other one).
I tested a couple of weeks a dataset of HIV Gag gene and there all tree methods (SLAC, FEL and REL) gave the same site under positive and negative selection.
When I test today the same dataset, the results of SLAC and FEL correspond, but the results of REL are opposite.
Please see the files in attachment to understand a little better what I mean.
It's driving me crazy that it won't work anymore and these results don't correspond.
Is it possible that there's a bug in the REL analysis (switching negative and positive)?

Kind regards
Kenny

---

Hi Kenny,

This does look very odd; I checked the results in HyPhy directly, and site assignments on Datamonkey are inverted. Let me investigate and post a solution here. Thanks very much for reporting the issue.

Sergei

Hi Kenny,

Sure enough -- there was a bug (introduced recently); to be more precise -- what should have been a bug (but used to work in the REL version on data monkey) is now functioning properly, but inverting the meaning of dS and dN (which is why positive selection became negative selection and vice versa). All fixed now, thanks to your timely report!

Best,
Sergei

Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login

Hi Sergei,

Many thanks for the quick solution and congrats for your excellent server  (Datamonkey) !

Kind regards
Kenny