I need to provide the number of haplotypes in my data for the hapFLK test (Fariello et al. 2013, Genetics).
I try to find the number of haplotypes in my whole-genome dog and wolf data (21Mb SNPs) with fastPHASE cross-validation procedure.
I created a test file (attached) and ran fastPHASE with different settings including:
./fastPHASE -T10 -KL3 -KU30 -Ki2 -Ks50 -Km1000 test.PHASE
However, fastPHASE always outputs K15, which I believe is incorrect because it never changes even if I change the input file. It also never outputs a file with _kselect extension, as it should. Maybe there is a bug. I use fastPHASE 1.4.8, it is available here http://scheet.org/software.html I tried to contact the author of the software, but have not got a response.
I also tried to use PHASE program, but it is extremely slow and cannot handle even a subset of 7K SNPs.
Could anyone share his/her experience with estimation of the number of haplotypes with fastPHASE or any other program?
Hello Dmytro,
Do you recode your SNP data before use them with fastphase ? In my code I have one line where I use plink --recode-fastphase on my data before fastphase.
Laure
Hi Laure,
Thanks for trying to help. Unfortunately, it was not an input issue.
After a lot of search online, I found out that fastPHASE 1.4.8 doesn't have the option to select K anymore https://www.biostars.org/p/214034/
I found this pipeline https://github.com/evodify/imputeqc which seems to work.
Kind regards,
Dmytro
For those, who will encounter this problem.
It seems the optimal solution is to use fastPHASE 1.4.0. It produces the _kselect file. However, fastPHASE 1.4.0 is not available for download anymore. If you use any computing cluster, v1.4.0 is likely to be installed there.
There is a solution for fastPHASE 1.4.8 too. It is described at https://github.com/evodify/imputeqc But according to my test, it is more laborious and less accurate.
According to the program developer, fastPHASE 1.4 doesn't support any more a cross-validation option. That is why I created a solution to estimate the optimal number of haplotype clusters for hapFLK with fastPHASE 1.4. It is available as an R package imputeqc -> https://github.com/inzilico/imputeqc
The instruction is here -> https://htmlpreview.github.io/?https://github.com/inzilico/imputeqc/blob/master/vignettes/k_selection.html
You can use fcGENE software (prepared for Win and Linux) to make fastPHASE input file from usual .ped and .map files .
- Badges
- Science method