View Full Version : Inquirie about IMPUTATION

09-08-2018, 10:06 AM
The guy from CNSGenomics on Github writes: "Imputation is a big, slow, ugly, long-winded, hand-wavey, unpleasant process. ". Indeed. I've myself learned with some difficulties to impute up to 3 millions SNPs or so to HAPMAP (CEU only) with PLINK, and this made I've been told that people never, absolutely NEVER, use PLINK as an imputation tool. Too bad for me. My question:
Does anyone on AG have a real and personal experience of a serious imputation pipeline, for example precisely the CNSGenomics ?https://github.com/CNSGenomics/impute-pipe/blob/master/README.md

09-12-2018, 09:04 PM
One more time it seems I opened an absolutely unuseful thread. I was a priori afraid by imputation, but in fact, it turns out that it's not more difficult than anything else. I used the version 5.0 of BEAGLE, and it worked like in a dream. My experience file was a 23&me file. I had to throw out the indels, and some multi-allelic loci, and in the resulting imputed vcf ( I made it for chrom 6 and chrom 1) I had to delete some other multi-allelic. I have now to download the other 1kg reference vcfs and with my bandwidth, it's not the least painful. But I am already in a position to say that I am virtually free from DNA.LAND.
Note: of course I don't intend to impute ancient genomes!