Order of commands in plink2

[Maryiam Shöâeè]

I have just run simple QC measures on some data that has been cleaned of heterozygous haploids etc..

Running 

--geno 0.1 

--maf 0.01

--mind 0.1 

on the data filters 97% of my samples out (383 out of 396 removed)

I don't think the data is so poor that it fails basic QC so badly. 

So I tried running each command by itself rather than in 1 starting with --maf, then --geno and finally --mind. I only lost one individual by doing this. 

Another version, i.e. running --maf first then running --geno and --mind on the new file  gives 

Error: All people removed due to missing genotype data (--mind).

I don't quite understand why this happens? Can someone explain what plink is actually doing that gives such dramatically different results when you change the order of the commands?

Thanks 

Maryam 

[Christopher Chang]

Hi,

Can you post the .log files for your runs?  Thanks.

[Maryiam Shöâeè]

Hi 

Sorry for the delay.

Here are the log files. 

Thanks for all your help

Maryam 

[Christopher Chang]

It looks like your dataset can be divided into two components: a small core of ~137k variants with nearly complete genotype information, and a collection of ~384k variants with a ~17% missing call rate.

By default, plink applies sample filters before variant filters.  (See https://www.cog-genomics.org/plink2/order for details.)  The average missing call rate in your dataset is around 13%, so "--mind 0.1" causes almost all your samples to be thrown out.

Manually running --geno first makes sense here.

[Maryiam Shöâeè]

Thanks for that! 

How did you figure out the missing rate of the 384k variants from the log files?? 

[Christopher Chang]

In addition to sample and variant counts, the log includes "Total genotyping rate" (i.e. fraction of non-missing calls).