Hi,

I'm building 4 taxon trees using fairly short alignments and I want to test whether one or more branches are under differing selection pressures. I have about 100,000 alignments to test and each is built from a single species' perspective.

I have therefore started by looking at using TestBranchDNDS.bf to test whether my starting sequence (which is my key interest) is under a different selection pressure than the rest of the tree. However I don't understand one of the options, the nsClass2 or nsClass1 rate parameter. What are these parameters and how will they affect the result? Also for those alignments where I get a significant result how do I tell if this branch is under positive or negative selection?

The alignments I am keen to locate from my dataset would hopefully show two of the species to be under negative selection (the genes are under constraint) while a third species has a more positive selection as the constraint has been released. I'm therefore weary of comparing my branch of interest against a global background if I suspect that background could comprise these different rates. Is there therefore a more appropriate test I could use?

I've read through the HyPhy book as well as other posts and papers and I can't seem to find an answer to my questions so I hope you can help,


Thanks,

Teri

Hi Teri,

I suggest you use a newer method (BranchSiteREL) to scan your alignment for evidence of episodic selection along any lineage. This method does not assume global background, and allows each branch to possess its own selective regime. Take a look at Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login and Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login for some background on the test.

Once you've run the 100,000 alignments through this analysis, you can screen those where the branch of interest is under strong positive selection.

Best,
Sergei

Sergei,

Thanks for the suggestion, so far it seems to be working but it is a little slow. How do I tell HYPHYMP to use multiple threads? I was expecting to add an argument when I start the program using my batchfile but I can't find any documentation about what to add,

Thanks for the help,

Teri

Hi Teri,

HYPHYMP (without any arguments) will try to use as many threads as possible (up to the # of CPUs in the system). For smaller datasets (4 sequences), the parallelization algorithm is not going to show much improvement, I am afraid. You may be better off compiling HYPHYMPI and using a wrapper file to run multiple fits in parallel.

Sergei