HyPhy message board
http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl
Methodology Questions >> How to >> how to do REL/FEL analyses
http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl?num=1114817177

Message started by dmerl on Apr 29th, 2005 at 4:26pm

Title: how to do REL/FEL analyses
Post by dmerl on Apr 29th, 2005 at 4:26pm
Hi Sergei - last question for today I promise :)

I am trying to use the REL and FEL methods, as described in KPF2005, and I just wanted to make sure that I have chosen the correct settings.  I am using the command line version of hyphy.  My understanding is that FEL corresponds to:

8 (positive selection) -> 6 (quickly test...) -> 1 (universal code) -> 2 (restore an anlysis...) -> 3 (estimate dn/ds from data with branch corrections) -> 6 (fixed effects, site by site, dS adjusted) -> ...

Is this a correct way to do a FEL analysis?

Also, is the Random effects likelihood method option 7 under the ancestor counting methods menu? (option 7 is "obtain a site specific distribution of dn and ds")  Or are your random effects methods implemented under the "test for positive selection using the method of Nielsen and Yang" menu?  

once again many thanks,
Dan


Title: Re: how to do REL/FEL analyses
Post by Sergei on Apr 29th, 2005 at 6:22pm
Dear Dan,

Firstly, my apologies for lack of documentation in the current HyPhy build. I will eventually write it up, I promise :P


wrote on Apr 29th, 2005 at 4:26pm:
Hi Sergei - last question for today I promise :)

8 (positive selection) -> 6 (quickly test...) -> 1 (universal code) -> 2 (restore an anlysis...) -> 3 (estimate dn/ds from data with branch corrections) -> 6 (fixed effects, site by site, dS adjusted) -> ...

Is this a correct way to do a FEL analysis?


This is correct, assuming that you are restoring an analysis file from an earlier (SLAC for example) run, otherwise the sequence changes to

-> 1 (new analysis...) -> choose model -> dN/dS option ->...



Quote:
Also, is the Random effects likelihood method option 7 under the ancestor counting methods menu? (option 7 is "obtain a site specific distribution of dn and ds")  Or are your random effects methods implemented under the "test for positive selection using the method of Nielsen and Yang" menu?  


The REL analysis can be replicated by:

2 (Codon Selection Analyses)
1 (Run an analysis; choose 2 here to process marginal files output by a REL analysis)
Choose data
1 (Univ code)
Tree
2 (Estimate branch lengths approximately)
Choose model
2 (Choose some models)
4 (Dual variable rates)
d (finish selection)
3 (Independent GDD)
1 (Use default parameters)
3 (syn classes)
3 (non-syn classes)
choose output file

When the analysis finishes it will write out a .marginals file in the place you specified at the prompt. You can then load that file using the '2' option in the 2nd step to get sites under selection etc.

The reason that REL is separate is that we actually have a separate MBE paper in revision on dual rate variation models. I am including a link to the manuscript (it should also make other analysis options more meaningful):
Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login

HTH,
Sergei

HyPhy message board » Powered by YaBB 2.5.2!
YaBB Forum Software © 2000-2024. All Rights Reserved.