These are chat archives for dereneaton/ipyrad

20th
Aug 2016
Katia Capel
@katiacapel
Aug 20 2016 20:42
This message was deleted
Katia Capel
@katiacapel
Aug 20 2016 20:56
Hi! I'm running ipyrad on my samples and there is something I can't understand. My first run included 8 samples, using the default parameters and all steps (1-7) I got a
total_prefiltered_loci of 61907. I wanted to see if I could get more loci reducing the number of samples to 6 (removing the most divergent and different ones), so I create a branch with a subset of samples and I ran only the steps 6-7 using the force flag (-f). I end up with less loci (total_prefiltered_loci of 44107). This patter keep happening, when I try to include only 3 samples highly similar with high sample coverage the total_prefiltered_loci is 20483 and I end up with no loci, all of are filtered in the process. Am I using the branching wrong??
Isaac Overcast
@isaacovercast
Aug 20 2016 22:08
Hi @katiacapel, You are actually using branching just right as far as I can tell. There are two questions here it sounds like, one is about total_prefiltered_loci going down as you reduce the # of samples and one is about the run with only 3 samples producing no output loci.
The second question is easier: The default setting for Min # of samples per locus is 4, so with only 3 samples you can see why this would output no loci. Change this setting in your params file and I'm sure you'll get output here
4                              ## [ 21 ] Min # samples per locus for output
set this to 3 and try again.
Isaac Overcast
@isaacovercast
Aug 20 2016 22:15
As for the first question: Greater numbers of samples will always (as far as I can tell) have greater numbers of prefiltered loci. This is because the total prefiltered loci records ALL loci for all samples (both shared among samples and private within samples). By removing samples from the analysis you reduce the total number of loci by the number of loci that are private within the excluded samples.
I hope I'm understanding your question right.