search

getian107 / PRScsx

Cross-population polygenic prediction
MIT License
69 stars 20 forks source link

How to get P value? #58

Closed YangRT11 closed 2 weeks ago

YangRT11 commented 2 weeks ago

Hi, I found that we can only get a posteriori beta values for our output file, can you please tell me how to get other effect values such as se and P, which is very useful for filtering of thresholds when constructing PRS.

getian107 commented 2 weeks ago

We do not recommend applying any filtering to posterior SNP weights. You can use all variants in the output to construct PRS. PRS-CSx is a Bayesian method and thus does not compute p-values for the posterior weights which is a frequentist approach.

YangRT11 commented 2 weeks ago

Thank you! May I ask if I want to construct a PRS using only some of the SNPs in the GWAS summary data (e.g. only SNPs with P<5E-8 are needed). Which of the following should I follow:

  1. Input file incorporating only SNPs with P<5E-8 and run PRS-CSx OR:
  2. Input file incorporates all SNPs from the GWAS summary data, and after running PRS-CSx, select the SNPs with P<5E-8 from the original GWAS from the output file, and further plink to build the PRS
getian107 commented 2 weeks ago

Bayesian PRS methods are developed to use genome-wide common variants. If you want to build a score using only genome-wide significant loci, you should check out the "clumping and thresholding" method and software. For example: https://choishingwan.github.io/PRSice/