Concordance factor to create "super-genes"

[Simon Uribe-Convers]

Dear all,

I have a large (500 loci) multilocus dataset with many semi-uninformative genes. I was thinking of running a Bayesian concordance analysis in BUCKy to identify loci that were congruent with each other. Then, I would take these congruent loci and concatenate them into "super-genes"—this way ending up with longer and, hopefully, more informative "genes" for a species tree analysis. Does this make any sense?

I feel like I've seen this done before but I haven't been able to find a citation for it. I would be grateful if someone can provide one.

Looking forward to your comments.

Best,

Simon

[Cecile Ane]

Hi Simon,
I can't find a reference right now but I seem to remember that Dave Weisrock and his group might have done that. If you want to group your loci, you're migt be better off using the "statistical binning" approach from Siavash Mirarab, Tandy Warnow et al. After you have your "super-genes", you can use your favorite approach to analyze them. 

If you wanted to use BUCKy, you don't necessarily need to build super-genes, though, because BUCKy integrates over gene tree uncertainty. 
Cécile

--
You received this message because you are subscribed to the Google Groups "BUCKy users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bucky-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Simon Uribe-Convers]

Hi Cécile, 

thanks for your response. My genes are UCEs and thus not very informative for BUCKy (or other) analyzes —that's why I wanted to create longer and more informative "super-genes".

The 'statistical binning' approach looks promising though, and it seems that the latest version (weighted statistical binning; Bayzid et al., 2015) concatenates genes based on gene tree "combinability", which is exactly what I wanted to do.

If you do think of a reference for doing it with BUCKy, i.e., concatenating genes that are concordant among themselves, please let me know!

Best,

Simon

[Cécile Ané]

I have not heard of anyone doing weighted statistical binning followed by BUCKy on the "supergenes", but Md. Shamsuzzoha Bayzid and Tandy Warnow looked at a number of species tree methods after "naive" binning in this paper: http://bioinformatics.oxfordjournals.org/content/29/18/2277.abstract . See the supplementary data for naive binning + BUCKy. I do *not* recommend naive binning though, because you might group genes with discordant genes in the same bin, unlike weighted statistical binning.

Cecile.

[Simon Uribe-Convers]

Thank you for all the recommendations Cécile! I think I'll go with weighted statistical binning followed by some summary species tree method.

Cheers,

Simon