HyPhy message board | |
http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl
HYPHY Package >> HyPhy feedback >> running parris with 2 synonymous rate classes http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl?num=1342224404 Message started by Dan Fulop on Jul 13th, 2012 at 5:06pm |
Title: running parris with 2 synonymous rate classes Post by Dan Fulop on Jul 13th, 2012 at 5:06pm
Hi HyPhy community,
I am trying to run PARRIS with 2 synonymous rate classes. What I've scripted to do this has two consecutive PARRIS runs: 1) a run with 2 synonymous classes, 2 non-synonymous classes, and Non-positive Discrete omega as the null model. 2) a run with 2 synonymous classes, 3 non-synonymous classes, and Independent Discrete omega as the alternative model. I will then calculate a LRT with 2 degrees of freedom on the above models. Is this the right way to run PARRIS with only 2 synonymous classes? Specifically, I am wondering if I need to specify different numbers of non-synonymous classes for the null and alternative models than what I'm currently specifying? Thanks in advance! Dan. |
Title: Re: running parris with 2 synonymous rate classes Post by konrad on Jul 16th, 2012 at 1:25pm
Dear Dan,
PARRIS was originally tested with Yang's M1a and M2a models for omega, i.e. the null has two non-positive omega categories of which one is fixed at omega=1 and the alternative has three omega categories of which one is non-positive, one fixed at omega=1 and one positive. The reason for the omega=1 categories is to prevent weak negative selection from contributing to a potential false positive signal, e.g. if the null had omega1=0.1 and omega2 = 0.4, then you might get significance for adding a category with omega3=1.05 even if many of the sites contributing to that really have omega=0.95. The number of synonymous classes shouldn't really be relevant - just needs to be a sufficiently good fit. I hope this helps, Konrad Scheffler |
Title: Re: running parris with 2 synonymous rate classes Post by Dan Fulop on Jul 29th, 2012 at 11:25pm
Hi Konrad,
Thanks so much for the informative response. That is precisely what I meant, i.e. how to constrain omega2=1 in the alternative model. I guess it just comes down to specifying omega2 as fixed at 1.0, which I don't know how to do in hyphy batch language. It would be great if you or someone else could point out how to fix the value of that rate class. Regarding the number of synonymous rate classes, the genes in my dataset have relatively little sequence divergence and I also have few taxa. So, using 2 instead of 3 synonymous classes should provide better power. Cheers, Dan. |
Title: Re: running parris with 2 synonymous rate classes Post by Dan Fulop on Aug 2nd, 2012 at 8:11pm
Hi Konrad and Sergei,
So, I just saw in an old command reference that the operator := will constrain a parameter prior to optimization. However, I just looked for over an hour through the PARRIS.bf code to see where I could constrain NSdistrInfo[0][1] := 1 for the 2 (syn) x 3 (omega) model (which is my alternative model), but I couldn't figure out. Any leads would be much appreciated! Thanks, Dan. |
HyPhy message board » Powered by YaBB 2.5.2! YaBB Forum Software © 2000-2024. All Rights Reserved. |