search

kauwelab / PolyRiskScore

PRSKB is a website and command-line interface tool for calculating polygenic risk scores using GWA studies from the NHGRI-EBI Catalog.
23 stars 1 forks source link

Comparison to PRSIce-2 #426

Closed erskck1 closed 1 year ago

erskck1 commented 1 year ago

Hi,

My question is about the (Comparison with PRSIce-2) section in the publication. In this section, you compared the distribution of both PRS results of PRSIce-2 and PRSKB. The question is, why didn't you report the correlation between the two results? It is clear that they have a identical distrubition but I would also like to know how was the correlation. Is there any way to know which parameters and SNPs you have entered into PRSIce-2?

Thanks and best regards, Ersoy

MattCloward commented 1 year ago

Thank you for your questions and I apologize that this information was not more readily available! I am looking for this information and will hopefully find it for you in the next few days.

MattCloward commented 1 year ago

I apologize for my extreme delay in answering your questions.

I have attached the log file of the PRSice2 analysis, which contains the settings used. I have replaced all paths with example paths instead. Specifically though, here is the command I ran:

Rscript "${PRSICE_DIR}/PRSice.R" --dir . --prsice "${PRSICE_DIR}/PRSice_linux" --base \"${assocFilePath}\" --target "${TARGET_DATA_SET}" --thread 4 --stat OR --binary-target T --no-regress --keep-ambig --out \"${OUT_DIR}/${targetDataName}_${assocFileName}\"

As for your other question, we chose to report a Welch’s two-sample t-test which accomplishes the same result as a correlation. However, here is the result of a Pearson correlation: Correlation = 0.9994441 p-value =<2.2e-16

PRSice2-GCST002245.log

erskck1 commented 1 year ago

Hi Matthew, Thank you very much for this issue too! It is clear to me now, we can close the issue. Best, Ersoy