Hello,

My sequences have 0.1%-2.5% diversity, is this too low to use SBP and GARD?  Other recombination tools say you need a minimum diversity (~5%).  I've noticed others publishing using GARD with similar data to mine though, and in the paper about these methods there were some low diversity alignments included, so I decided to give it a try!

I ran sequences from the person with the highest diversity and SBP resulted in Yes for recombination (cAIC 4.25 improvement, 89.7% support), then the GARD said no recomb (used the model selection tool and 4 rate categories).  I'm very new at this, is it normal for the 2 to disagree?  My plan was to use SBP to determine if there is recombination, then follow-up with GARD to find the breakpoints.  Are the differences in my alignments too few to expect findings of recombination?  I did notice in table 1 of  the paper where delta was <100 there were no breakpoints.

The sequences are ~1000 bp and ~15 per alignment.

Thank you!
Liz  

Hi Liz,

A c-AIC difference of 4.25 is borderline (89.7%) significant, and I would say in this case insufficient to support recombination. It is possible to find strong recombination signal in a low diversity sample (assuming the fragments are long enough).

If I recall correctly, the breakpoints in the GARD paper were reported in they were confirmed by another (Kishino-Hasegawa) test for phylogenetic incongruence given the breakpoints.

Cheers,
Sergei

Hi Sergei,

Thank you!  I have one additional question (for now!!).  I read some of the other posts, and am trying to use HyPhy GARDProcessor.bf

I uploaded my alignment, and the [CSV] output from GARD.  It's been running on my (old) computer for 3 days, and it's still fitting the single tree (currently tree_300).  Did I do something wrong, is is caught in a strange loop, or is it just my horribly slow computer?  

I know this is slightly off-topic, but if I need to seek out a faster computer I figure it's best to find out sooner, I've got a couple of these to do, and a deadline  :o.
Thank you,
Liz

Hi Liz,

You no longer need to run GARDProcessor on your machine. KH tests and other metrics are now reported on the datamonkey.org results page.

Sergei

Thank you!  I was pretty confused as to why there was a KH report at the bottom of GARD if we had to run it separately, but I'm still at the following directions stage, trying to move into actual understanding!!

Liz