GWAS & PRS tutorial

[Sintia Belangero]

Hi all, 

I have wondered whether there is a plink2 standardized script for GWAS and PRS or another option for getting set up. That will be super helpful for grad student with starting to analyzing, a step-by-step tutorial considering QC and PC/ ancestry steps. 

Thank you in advance. 

Sintia Belangero 

--

-----------------------------------

  Sintia I. Belangero

  -Associate Professor, Department of Morphology and Genetics 

  -Coordinator of The Lab of Integrative Neuroscience (LINC)

  Universidade Federal de São Paulo

  São Paulo-SP / Brazil

  https://www.unifesp.br

  http://pgtunifesp.com.br

[Christopher Chang]

For a plink-and-PRS tutorial that includes a PCA step, see https://choishingwan.github.io/PRS-Tutorial/plink/ .  With the exception of --clump, all steps can currently be performed with plink 2.0 instead of 1.9.

From there, simple GWAS can be done with

plink2 \

  --bfile <input fileset prefix> \

  --glm \

  --pheno <phenotype file> \

  --covar <.eigenvec file from --pca> \

  --out <output prefix>

Let me know if you need more assistance.

[Yuyang Luo]

Hi,

I am wondering how to extract the residual for GWAS results using plink1.9 or plink2. Thanks!

Best,

Yuyang

[Christopher Chang]

If you want every single residual value, see https://groups.google.com/g/plink2-users/c/fNNGzvEXC58/m/u5KxpwGTBAAJ .  It doesn't really make sense for plink2 to report these values directly, since the output file would often be gigantic.

If you want a more narrow statistic, can you clarify exactly what you want and what you are using it for?

[Yuyang Luo]

Thanks! 

As you mentioned, what I would like to calculate is the difference between the estimated phenotype value (which you compute from the regression coefficients, genotype value, and covariates) and the actually-observed phenotype value.

I would to check the normal distribution of residuals for our GWAS. Can we calculate it based on the GWAS summary report? Thanks.

Best,

Yuyang

[Yuyang Luo]

I am wondering how to calculate estimated phenotype data for each subject based on GWAS summary table. Thanks

[Christopher Chang]

No, you can't calculate the entire distribution of residuals from the GWAS summary report alone.

However, if you run --linear/--logistic/--glm with the "intercept" modifier, you can load the resulting coefficients and the genotype/covariate data, compute the predicted phenotype values, and subtract them from the actual phenotype values.