"--glm cols= +a1freq" calculate MAF not a1 frequency

[SungJi Moon]

Hi

I'm trying to use column set descriptor "a1freq" in plink2 --glm operation like followings

plink2 --bpfile myfile --glm cols=+a1freq ..

But, output of a1freq is not a1 frequency but minor allele frequency which is one of a1 frequencey or ref allele frequency.

It is same in glm.linear, glm.logistic output (case/control)

I'd appreciate it if you could check it out

Many thanks,

Sungji

[Christopher Chang]

This is working exactly as designed.  Why do you think ALT and A1 are distinct columns in the .glm.linear and .glm.logistic reports?  If they couldn't be different, there would be no point...

In plink2, "A1" always means an allele that DOESN'T have to be the first ALT allele; otherwise it would have just been named "ALT"/"ALT1".  For --glm, A1 is the allele that the current regression corresponds to; for each variant, this loops through all but one of the alleles (one allele must be omitted to avoid linear dependency).  Use --glm's "omit-ref" modifier to make --glm always omit the REF allele, instead of following plink 1.x's default behavior of omitting the major allele; and be aware that this will make interaction tests report "NA" results for no good reason when the REF frequency is too low.

[SungJi Moon]

Thank you for concrete answer.

I have one more question.

In plink2 glm operation, A1 allele is test allele.

Then, is it right understanding that among REF allele and ALT allele, minor allele (lower frequency) becomes A1 allele (test allele) ?

and if I add "omit-ref" modifier, then ALT allele becomes A1 allele (test allele) ?

Thanks,

Sungji

2019년 10월 20일 일요일 오후 11시 57분 8초 UTC+9, Christopher Chang 님의 말:

[Christopher Chang]

That is correct.