HyPhy message board | |
http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl
HYPHY Package >> HyPhy feedback >> BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl?num=1181946545 Message started by matty on Jun 15th, 2007 at 3:29pm |
Title: BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! Post by matty on Jun 15th, 2007 at 3:29pm
Hi Sergei,
I have just run the YangNielsenBranchSite2005.bf test, and I have found several sites under positive selection with Bayes Factors well over 100. I have also run the branch site model implemented in PAML, specifically Test 2. PAML returns the same sites as HYPHY with significant values for Naive Empirical Bayes (NEB), but returns no positively selected sites with Bayes Empirical Bayes (BEB). The authors of PAML recommend ignoring the NEB values, thus I am wondering which results are correct. Should I follow Yang's advice and ignore NEB sites, and thus disregard the HYPHY results? Or should I use the results returned by HYPHY? I'm thinking that it could be that the BEB was not powerful enough to pick up any sites, even though positive selection did occur (given a significant LRT and an omega > 1). What do you believe? Thanks for your time, Matt |
Title: Re: BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! Post by Sergei on Jun 15th, 2007 at 4:10pm
Dear Matt,
BEB is really just a hack to attempt to approximate the results of a fully Bayesian inference without actually doing it; it allows one to incorporate sampling errors in parameter estimates. Generally speaking, if your data set is small and parameter estimate errors are lare (the most likely scenario where NEB and BEB might disagree), there is very little power to detect individual sites under selection, regardless of which method you use. The fact that LRT is significant (for omega>1) does not guarantee that you will be able to find at least one site under selection, e.g. the LRT could be due to small contributions from a number of sites, but individually, no single site is reliably significant. If you are really interested in pinning down the significance of your findings, do some parametric simulations; generate 100 data sets using the omega distribution (and all other parameters, such as base frequencies and branch lengths) inferred from your alignment, and see what type of false positive error in detecting individual sites under selection you get. Then you can tweak the bayes factor cutoff to make that small. Cheers, Sergei |
Title: Re: BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! Post by matty on Jun 20th, 2007 at 10:10am
Hi Sergei,
Thanks for your response. I am familiar with how to do the parametric simulations you suggested using the likelihood functions created via the data panel interface (I performed this on one of your example datasets). However, I cannot get it to work for a standard analysis on my own dataset. It complains that I cannot save a likelihood function if i did not use the data panel interface (ie. I just loaded in my data). Any help would be appreciated. Matt |
Title: Re: BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! Post by matty on Jun 20th, 2007 at 5:08pm
Hey Sergei,
Thank you for all your help. So I'm in the process of implementing what was stated, but I've hit a little snag. When it comes time to choose the node to set as the foreground, I'm having trouble when listing my choices of nodes as stdinRedirect[]. I would like to find a way to let the program know that there are no more nodes I would like to select. When running this on command line without a batchfile, I would normally hit "d", but it does not accept this as part of my stdin array. I found that if I make it implicit in choiceList the amount of nodes I want to choose, then the problem is solved. But then if I only want to only choose ONE node as my foreground, I get an error stating "Operation MAccess is not defined for 0" Do you have any idea on how to get around this? Thank you again for your help! Matt |
Title: Re: BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! Post by Sergei on Jun 20th, 2007 at 5:39pm
Dear Matt,
Instead of "d", place an empty string "" in the stdinArray structure to terminate the variable length selection. HTH, Sergei |
Title: Re: BEB, NEB and Bayes Factor...Ahhhhhhhhhhh!!! Post by Andrew_Roth on Jun 20th, 2007 at 10:25pm
Hi Sergei,
I am working with matty on this project. Thanks for all the help so far, we have got our batch script to run and we are just waiting on our server to finish analysing the simulated data sets. The script is setup to save the likelihood function from each dataset. I was wondering if there where any articles or books you could point us towards that might help us understand how to make use the bootstrap data. I have done some reading online but all the articles seem to be about using bootstrapping to test the robustness of inferred trees. We are looking to find out if the positively selected sites we have found are indeed positively selected. They tend to have very high Bayes factors anywhere from 70 - 40000. As per your suggestion we would like to use the bootstrapping procedure to find a reasonable cutoff for the Bayes factor but I am unsure of how to go about this. Thanks again for all the help. Cheers, Andrew |
HyPhy message board » Powered by YaBB 2.5.2! YaBB Forum Software © 2000-2024. All Rights Reserved. |