2cjenn / prs_pipeline Goto Github PK

• Had assumed loading via module system, can't assume this, not all users will have this type of HPC system
• Had assumed running via script, but need commands to be suitable for running live (or have live variants)

<span class="btn btn-primary glyphicon glyphicon-copy"></span>

That makes a nicer button

See RLesur/klippy#24

Hi,

Thank you for the wonderful step-by-step explanation of how to construct PRS in the UKB set.

I was trying to follow the document and I was not sure about the section 'Multi-allelic SNPs and alignment' and in particular the paragraph re alignment which is

"When doing this we need to take into account the alignment of the SNPs. We know that the UK Biobank data is aligned to GrCh37, with the first allele in the bgen files (‘allele1’ column) the reference allele on the forwards strand. Our betas file, on the other hand, has the alleles labelled as effect and noneffect - and this labelling might be determined by which allele is more common, or has a particular direction of assocation with the trait of interest.

For this reason, when we do our join in SQL, we can’t just assume that all our effect alleles will be in the same column of the index file. To account for this, we use two join statements, one that matches up variants where ‘allele1’ is our noneffect allele (and ‘allele2’ is our effect), and one with the opposite alignment. Notice that we flip the sign of the beta on the second join, as discussed in Section 2.4.1 of our companion paper".

I am slightly confused regarding a situation that you have mentioned in your paper (Row 2, Table 4; A/C in the PRS summary data and C/A in the validation data). In the paper you say that it can be resolved automatically by some softwares such as PLINK2 but in the tutorial you say that you have to flip the sign of beta when joining the summary data and validation datasets.

So my question is:

1. Does it mean that for a situation where effect and non-effect allele are inverted between datasets you don't have to flip the sign of betas if you are using PLINK2? I could not find a reference for this in the plink manual.

Thanks.

• Contains quotation marks when downloaded (not visible in Excel)
• Column names don't match those used later

Hi, Thanks for your works.
I am happy to see the pipeline. But i have to figure out one thing that some files may not get from UKB. I can access the ukb. But i can not find the file (ukb_mfi_chr${i}_v3.txt). We may not get the files again.
Or can you provide the way to get the files? Thanks.