This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Info Score > 1 Plink?

What does an info score > 1 mean? I thought all scores should be between 0 and 1. Info score comes from assoc.dosage file in plink.

plink gwas score
• 5,244 views • link
updated by chrchang523 chrchang523 • written by bg1116 • bg1116

did you check the plink manual pages:

The fields INFO and NPRX refer to how well PLINK managed, if at all, to impute the SNP. If NPRX is zero, then it could not be even poorly imputed. If INFO ranges from between 0 and 1, although it can be greater than 1 occasionally. A higher value general means a better imputed SNP; roughly speaking, only looking at imputed SNPs with a INFO value greater than 0.8 or so is probably good practice. More specific details on these metrics will be posted soon.

...I put in bold the important line.

Kevin

• 0 views • link
written by Kevin Blighe Kevin Blighe

INFO has the same meaning in both files, but that documentation snippet is actually for a slightly different (and defunct) command, --proxy-assoc instead of --dosage.

• 0 views • link
written by chrchang523 chrchang523

1 answer

The INFO field in an .assoc.dosage file contains the MaCH r-squared imputation quality metric, which is [observed dosage variance] / [expected variance, given observed allele frequencies and Hardy-Weinberg equilibrium]. Despite the r-squared name, this is not guaranteed to be less than or equal to 1, especially when there's significant deviation from Hardy-Weinberg equilibrium; but low values can still be trusted to correspond to poorly-imputed SNPs.

• 0 views • link
written by chrchang523 chrchang523

When performing association tests on dosage data, it looks like my poorly imputed variants (Rsq < .3ish) get NA pvalues. I would usually filter those out anyways, but out of curiosity is there an internal cutoff in PLINK that excludes sites with low Rsq from the test?

• 0 views • link
written by curious curious

Log in to answer this question.

More posts like this