HyPhy message board - All codons show selection

cdm

YaBB Newbies

Offline

Feed your monkey!

Posts: 2

All codons show selection
Mar 3^rd, 2010 at 3:06pm

I've been testing for selection at 3 MHC loci (89 codons) in ~10 species, using REL. I have between 5-25 alleles generally per species, and most show reasonable results; 0-10 codons show positive selection.
However at two loci in one of the species, huge numbers of codons (88/89) and (38/89) show positive selection, whilst just 9 and 16 codons are actually variable. I know REL has false positive issues, but this seems incredibly high. Is this a bug, or small sample size issues (15-16 alleles)? FEL suggest just a few sites with selection.

Back to top

IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: All codons show selection
Reply #1 - Mar 3^rd, 2010 at 3:10pm

Hi cdm,

cdm wrote on Mar 3^rd, 2010 at 3:06pm:

I've been testing for selection at 3 MHC loci (89 codons) in ~10 species, using REL. I have between 5-25 alleles generally per species, and most show reasonable results; 0-10 codons show positive selection.
However at two loci in one of the species, huge numbers of codons (88/89) and (38/89) show positive selection, whilst just 9 and 16 codons are actually variable. I know REL has false positive issues, but this seems incredibly high. Is this a bug, or small sample size issues (15-16 alleles)? FEL suggest just a few sites with selection.

There are rare cases (e.g. few sequences with only non-synonymous variation), where REL gives such results. They are completely unreliable (false positives) and should be ignored, especially if not confirmed by other methods.

Sergei

Back to top

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

cdm YaBB Newbies Offline Feed your monkey! Posts: 2	Re: All codons show selection Reply #2 - Mar 3^rd, 2010 at 3:58pm The other species I tested that appear also had few sequences and dN>>>dS, but show more reasonable results - are these valid or do I ignore them too? REL is presumably always likely to suggest more sites with selection than FEL because it is less conservative.
Back to top	IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: All codons show selection
Reply #3 - Mar 3^rd, 2010 at 4:01pm

Hi cdm,

cdm wrote on Mar 3^rd, 2010 at 3:58pm:

The other species I tested that appear also had few sequences and dN>>>dS, but show more reasonable results - are these valid or do I ignore them too? REL is presumably always likely to suggest more sites with selection than FEL because it is less conservative.

Generally, speaking there is very little power to detect any individual site under selection in a small/sparse sample, hence I would advise against placing too much weight on REL inferences in this case. Run PARRIS to confirm the presence of selection in the gene as a whole as another checkpoint.

Sergei

Back to top

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

	Welcome, Guest. Please Login