plink case/control association analysis and Manhattan plot

Question

[nayeona.hi]

There is no one to ask about the analysis process, please advise to me.

I'd like to do perform association test between case and control group.
First of all, I made 'phenotype.txt' coded as 0, 1.
As I'm not sure, I did 3 kind of association works for all samples, case samples, and control samples.

1. ./plink --noweb --bfile <all data> --assoc --pheno <phenotype.txt> --out <new file> 
2. /plink --noweb --bfile <case data> --assoc --pheno <phenotype.txt> --out <new file>
3. /plink --noweb --bfile <control data> --assoc --pheno <phenotype.txt> --out <new file>

(maybe of course,) the top SNPs on each Manhattan plot were different, likewise on p-value.
Then, can I choose the specific SNPs from the case data for the phenotype, and understand the SNPs might affect the phenotype?

Or, since the process has a problem, do I have to do other methods?
Making a new PED file coded 'affection' column as one of 'phenotype.txt'?

Please help and hope to advice. Thank you.

[brunobsouzaa]

For this kind of analysis, there's a few steps that you have to do (just checking what you've done so far!). Firts perform test for HWE then perform IBD. For your affection, plink will assume that this info is in a specific column of your ped file (check the manual). Just code it as 0 and 1 and you're good to go. To perform statistical analysis, check which test better suites for your data. You can perform a simple chi-square, Fisher exact test or logistic regression. Just tell plink which test you want to perform. It's not necessary to create separeted files for you cases and controls group!

Hope it helps.

[nayeona.hi]

you mean I have to edit the PED file (or bim file instead?) with coding 'affection column', right? Thank you and I will try.