PLINK linear association testing command with an interaction and 2 covariates term

Question

[entropy]

I am trying to write the right command for my association testing using PLINK 1.9 but I could not find a clear answer so far and not sure how to wrtie it correct.

I have genotype variants called and in PLINK bed formatted files under geno_data folder.

I have phenotype.txt file. First column as "Sample_ID", second column as "Treatment" (1 for Placebo and 2 for CompoundX) and third column has the Response variable as "ResX", which has integer number between 0 to 9.

I also have PCs file as cov.txt. The first and second columns are the same and includes Sample_IDs. The column names are "FID" and "IID". The third, fouth and fifth columns have PCs,with the column names "PC1", "PC2" and "PC3".

The Plink documentation did not help me as I am new in this. Therefore I need some examples command to clarify for various models as described below.

I need the correct PLINK (1.9) command (if not possible PLINK2 is also OK) for the following linear models:

Model 1:

ResX ~ Variant * Treatment + PC1 + PC2 + PC3

Second command I need is for this model:

Model 2:

ResX ~ Variant + Treatment + PC1 + PC2 + PC3

Model 3:

ResX ~ Variant * Treatment

Thank you

Answer 1

[chrchang523]

See the "Flexibly specifying the model" section under https://zzz.bwh.harvard.edu/plink/anal.shtml#glm , discussing --interaction and --parameters.

[entropy]

I have read the documentation many times but not fully clear to me. Simple examples always clarify. I had run the below command and it seems worked with PLINK2 at least.

I made this pheno_cov.txt file first with these headers and some additional unrelavant columns in it as well.

FID IID ResX Treatment PC1 PC2 PC3

Then run the below PLIN2 command for the model 1:

plink2 \

--bfile geno_data/mydata \

--pheno pheno_cov.txt \

--pheno-name ResX \

--covar pheno_cov.txt \

--covar-name Treatment,PC1,PC2,PC3 \

--parameters 1,2,3,4,5,6 \

--glm interaction

[chrchang523]

Examples corresponding to your use case are already present in the part of the documentation I linked.

You need to clearly show how you are attempting to adapt the main example to your use case, and how that is failing.

[entropy]

In that link this this command seems relevant but not fully clear to me:

plink --bfile mydata --linear --covar tmp.cov --interaction --parameters 1,2,3,5

Because, in this example the tmp.cov header informtion is not given or I cannot see it there. I am not fully sure how to accurately write the paramteres without seeing the tmp.cov.

In my case the tmp.cov file has the pheno_cov.txt file and I had given the header information with 7 column names as:

FID IID ResX Treatment PC1 PC2 PC3

After trial and error, it looks the pramateters started counting from the third column, which is ResX in my file, and therefore the parameter 6 of "--parameters 1,2,3,4,5,6 "generated me the ADDxTreatment TEST result.

My goal was to see the effect of Treatment with SNPs on the ResX response variable by incorporating the PC1, PC2 and PC3 as only covariates for population stratification purpose.

[chrchang523]

Ok, let's dig deeper into your current Model 1 command line.

Look at the top 7 lines of the main output file, which should include the header line and all results for the first SNP. You can see an ADDxTreatment term there; but do the rest of the terms correspond to what you want in the model, and only what you want?

[entropy]

Ther terms in the TEST column are ADD,Treatment, PC1, PC2, PC3, ADDxTreatment. It does not seem to have only what I want. Does including them make any difference on the ADDxTreatment result? It looks like a better command might be like "--parameters 1,6 ":