Welcome, Guest. Please Login
YaBB - Yet another Bulletin Board
 
  HomeHelpSearchLogin  
 
Codon-based selection analysis for homologous sequences and GARD (Read 683 times)
Jack wong
YaBB Newbies
*
Offline


Curious HyPhy user

Posts: 1
Codon-based selection analysis for homologous sequences and GARD
Sep 14th, 2012 at 8:48pm
 
Hi,

I am new to the forum and also new to use the selection methods so have some basic question regarding analysis. Would be pleased to have opinion and suggestions!

1. Is codon based analysis suitable for within species(homologous genes) selection analysis? The reason being high recombination within the population?

2.I used GARD method and the program gave me one potential break-point with High AIC score with exact location. However, the program out put read like this "The alignment contained 85 potential breakpoints, translating into the search space of 3655 models with up to 2 breakpoints of which 79.15% was explored by the genetic algorithm".
Bit confused now how does 85 potential breakpoints leads ultimately to one. Cheesy


3.For a particular codon, selection analysis using GARD inferred tree I could only get positive selection from two methods REL(Bayes factor 190.59) and FUBAR(posterior probabbility 0.915). How reliable are these two significance values?

Thank you in advance! Smiley

Jack









Back to top
 
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: Codon-based selection analysis for homologous sequences and GARD
Reply #1 - Sep 17th, 2012 at 4:42pm
 
Hi Jack,

To answer your questions (if you are looking for more documentation, look at Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login and Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login

1. Generally speaking you should be OK, so long as you use a partitioning analysis. When writing the GARD paper we found that this correction works well even for close to 'free recombination' scenarios.
2. "85 potential breakpoints" refers to how many positions in the alignment could have been tested for a presence of recombination breakpoint. This statistic tells you how many variable sites you have in your alignment (essentially), and how many combinations of 1 and 2 out of 85 there are (3655). This is more of a diagnostic of how complex the problem is, not of what the solutions are.
3. See the first above. Reliable is a relative term . Posterior probabilities of as low as 0.5 have been used to detect selection in other applications, but 0.9 or 0.95 is more customary. Datamonkey defaults are 20 for Bayes factors and 0.9 for posterior probabilities.

Sergei
Back to top
 

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