HyPhy message board
Theoretical questions >> Sequence Analysis >> GARD

Message started by Kartik on Apr 12th, 2011 at 11:23am

Title: GARD
Post by Kartik on Apr 12th, 2011 at 11:23am

First of all, my apologies if this is a silly question. I'm new to phylogenetics. When I run GARD and SPB analyses on datamonkey server, GARD tells me that it did not find any evidence for recombination. SPB on the other hand identifies '503' as a possible position for recombination. I want to know...

1. Am I interpreting the results wrongly? Are the two algorithms used for different purposes?

2. If SPB is correct and if there is a breakpoint at the particular position, then will it have an impact on the selection analyses (site models and branch-site models)?

3. If it does have an impact, then can I overcome this problem by dividing the alignment into two partitions (1-503 and the rest) and carryout the selection analyses?

Thank you very much,

PS: Datamonkey and hyphy rock!


The alignment contained 430 potential breakpoints, translating into the search space of 430 models with up to 1 breakpoints of which 350.93% was explored by the genetic algorithm.

Recombination report

GARD found no evidence of recombination


AIC      Yes      46.3      503      100.00%

Title: Re: GARD
Post by Sergei on Apr 13th, 2011 at 1:44pm
Hi kartiks,

GARD uses a more conservative information criterion (small sample AIC, or AIC-c); it requires stronger evidence to declare recombination in an alignment. In your case, SBP should have also found no recombination under AIC-c or BIC. The original manuscript conducted simulations and other comparative studies using AIC-c; AIC is reported by SBP for reference purposes only.

In terms of correcting for the confounding effect recombination, please refer to section 3.10 of Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login for references and discussion.


Title: Re: GARD
Post by Kartik Sunagar on Apr 13th, 2011 at 2:14pm
Dear Sir,

Thank you very much for the reply.

When you say SBP results under AIC should be used just for reference, does it mean I should consider the results of GARD? That there is no recombination in the dataset?

Thank you very much for your time,

Yours Sincerely,
Kartik  :)

Title: Re: GARD
Post by Sergei on Apr 13th, 2011 at 4:29pm
Hi Kartik,

Yes, please use the results of GARD and SBP under AIC-c. From what I recall from my simulations in 2005, AIC gives a lot of false positives, and BIC -- too many false negatives, hence they are not recommended as the default criterion.


Title: Re: GARD
Post by Kartik Sunagar on Apr 14th, 2011 at 1:34pm
Thank you very much Sir, for the reply :)

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