NaN values in some mQTLs parquet file

[ariadnacilleros]

Hi,

I have been conducting several nominal mappings with TensorQTL, and I have realized that, in all the chromosomes, there are a few mQTLs for which I got NaN statistics (p-value, slope and se). The variants don't have rare MAFs or a smaller amount of counts, do you have any idea why this could be happening? Here I paste a few of them.

Thanks!

[francois-a]

Hi, it could be due to missing values. Otherwise, are you able to share the inputs for these specific examples?

[ariadnacilleros]

Hi,

I am sorry, but I am not able to share the input data, I only have the summary stats. But it seems to me that the problem is the SNP, although I am not sure if this would be due to missing data, as ma_samples and ma_counts are correct, are you suggesting that these SNP could be coded as NaN in some samples?

Thanks!

[Dongmeng-wang]

Hi,

I also got NaN for phenotypes with NA. Is it possible that tensorQTL can deal with the missing values of phenotypes? Thank you so much!

All the best,
Diana

[francois-a]

Without more information to reproduce this, the most likely reason for NaNs in the output are missing values. Missing genotypes are imputed to the mean, but missing phenotypes are not imputed since the best imputation strategy is likely data-specific. You'll need to impute/remove any missing phenotypes before running tensorQTL.

[ariadnacilleros]

Hi François,

Thank you so much for the answer, we will take this into count.

Best,

Ariadna Cilleros