Workflow clarification

[eyu8]

Hello,

Love the methods implemented in PTWAS and the reasoning behind them.

There are some parts that are unclear to me about the workflow.

1. Using the significant weights you provided, it is not necessary to run ptwas_est, if I understood correctly, as it is used when we implement our own eQTL weights?
2. Am I correct that the example GWAS summary statistics provided and the 1K reference are aligned with hg19 while the weights are hg38? Will it work with GWAS summary statistics from hg38?
3. Formating the GWAS summary statistics with CHR, POS, ALT, REF, ZSCORE ended with a segmentation fault for me. Only when I removed the header and changed the columns to CHR (as chr1, not 1), POS, ALT, REF, SNP_ID (CHR and POS combined with colon), N, and ZSCORE. I suggest that this should be looked at as the format for the GIANT summary statistics was the correct one and not the one in your documentation.

Thanks,
Eric

[xqwen]

Hi Eric,

1. We suggest run PTWAS testing procedure before the estimation procedure (ptwas_est). There is indeed a separation between testing the potential causal relationship from gene->trait vs. estimating the magnitude of the causal effect. You certainly can stop at the testing stage just as most available TWAS methods do.

2. I need to check with the authors of GMABIT and get back to you on that issue. We should provide a hg38 panel.
   We actually recommend using the reference panel reflecting your eQTL samples. In the paper, our panel is built based on the GTEx samples. But we can't release those panels because of privacy concerns. If you have dbGAP proved access to GTEx data, send an email to me.

3. Thanks for letting me. We will check with the GAMBIT authors and correct the documentation.

Thanks again!

William

[eyu8]

Thanks for the prompt response!
One last thing, the *.summary_out.txt files created by GAMBIT seems to be missing the header for the last column, where I believe it should be INFO.

Looking forward to the changes,
Eric