Large differences in eQTLs between GTEx v7 and v8

Question

[pcmph]

Can anyone explain why there is such a huge difference in eQTLs between GTEx v7 and v8? Specifically, I am looking at lung-specific eQTLs. After filtering by pval_beta < 0.05, assigning variant_ids to rsIDs, and ensemble IDs to gene names, comparison of gene-rsID pairs showed the following:

overlap=intersect(v7,v8)

length(v7): 1,462,393

length(v8): 2,379,279

length(overlap): 297,222

This suggests only 20% of the lung eQTLs in v7 are found in v8.

Any input would be much appreciated. Thank you.

Answer 1

[GouthamAtla]

A direct overlap is not the way to assess QTL replication. Both version might have used slightly different sample sizes and they used different genome versions (hg19 vs hg38) and probably slightly different methods. So the lead QTL for each gene might be different in both versions.

The systematic way is to measure "π1 statistic" to quantify the replication.