3
$\begingroup$

this is a very basic question, but I cannot find it explicitly stated anywhere: When exactly should I filter the info scores after imputation?

I did imputation in Impute2 and am planning on using SNPtest to analyze, but cannot tell if I should filter the scores in Impute2, SNPtest, Gtool, QCtool, or through python.

I'm hoping to get rid of the poorly imputed results (with an info score cutoff <0.5) in a population of ~4000 (half cases, half controls). After this, I'm running an association analysis -- probably through PLINK, but I'm still figuring out the best way to go from Impute2 to PLINK.

Any help would be greatly appreciated!!

Improve this question
$\endgroup$
0

2 Answers 2

Reset to default
2
$\begingroup$

In simple GWAS setups, each SNP is analyzed independently. In those cases you can filter out SNPs with poor INFO scores at any point.

For analyses that combine information across SNPs (for example FaSTLMM), I would recommend filtering out SNPs before running the step that combines information across SNPs.

So, if you filter right after running Impute2, you should be safe.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks so much, winni2k. What is the best way to filter the SNPs? Is that usually an option when inputing files to a software, or something that should be done separately (e.g., through something like python)? $\endgroup$
    – hp2018
    Feb 23, 2018 at 18:58
  • $\begingroup$ Plink is wildly popular and very fast. But do use version 2 if you want to avoid surprise allele reordering. $\endgroup$
    – winni2k
    Mar 1, 2018 at 7:57
1
$\begingroup$

Depending on the size of your samples (1K, 10K, 100K?) or the size of imputation region (1 gene, 1 chromosome, whole genome?) imputation files even after zipping are usually huge.

Instead of filtering SNPs and creating yet more gazillions of subset files. I prefer to have excludeSNP.txt file list of SNPs. This list then used as input for further analysis using SNPTEST flag -exclude_snps_g. This list is part of IMPUTE2 output or could be additional list of SNPs that we wish to exclude for other reasons. In short, filter at the point of analysis not the imputated files.

If info file is missing we can run SNPTEST with -summary_stats_only flag, which gives you the info score.

Here is an old still relevant post at BioStars post-imputation QC.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.