Handling z-score in sumstat

[jiyue1214]

Find a solution to handling the z-score in sumstat and implement it in the harmonisation pipeline.

[jiyue1214]

Our current harmonization pipeline cannot process the z-score column in the summary statistic data. However, the z-score is one of the outputs from METAL, so we’ve decided to address the z-score for flipped alleles in the harmonization pipeline.

After reading a blog about the relationship between z-score and beta (OR), it seems that when we flip alleles, the new z-score is equal to the negative of the original z-score.

Karatug and Trang from the AMP gave me the green light. Sam had the idea to convert the z-score to beta and do what we normally do. However, when the z-score is available, see is not available.

Should we go with new z-score = -(old z-score)?

[jiyue1214]

In the GWAS-SSF, we did not define the suggested header name of the Z-score. Investigating how the Z-score is presented among the current GWAS data we have:
Zscore (METAL), zscore, z-score,z,Z

1595 studies containing '\t'z:
1179 z
1 z.meta
44 z_score
3 zscore
1 zscore_metal
26 z_value
Even though the z is more commonly used as the header for the Z-score, personally think it is too short and not informative enough. I am using the z_score in the pipeline, but it is easy to change.