Empty trans-eqtl results

[vivekruhela]

Hi,

I am trying to perform the cis and trans eqtl analysis. I have two cohorts. One cohort (cohort-A) has 25 samples and another cohort (cohort-B) has 250 samples. For cohort-B, both cis and trans eqtl is working fine. But for cohort-A, only cis is working fine and trans eqtl is generating empty results. I tried tuning the parameters such as batch size, p-value, maf, etc. but in all the case, it is giving my empty trans eqtl results. Also, when I merge cohort-A and cohort-B to make a one large cohort, both cis and trans eqtl are working fine. I am curious to know that is it related to the batch size??

[francois-a]

Can you share a reproducible example? Batch size only controls how much of the calculation is done per iteration (and fits into memory) and doesn't affect the results.

[vivekruhela]

Hi,

Unfortunately, I can't share the data because it is not published. But I think this issue can be checked on any example data by reducing the number of samples. (e.g. check with 10/20 samples only)

UPDATE: For the same cohort I mentioned previously (Cohort-A), now I have 32 samples and both cis and trans eqtl are working fine.

[francois-a]

Try using the --return_dense option, and set --maf_threshold 0. Without seeing the data the only explanation is that no variants pass the filters.