Missing values ImputedDosageMatrix.dat

[harmbrugge]

When there are missing values (-1) in ImputedDosageMatrix.dat, no errors are given when running the eQTL pipeline.
But more importantly the direction of the effect sizes are not always correct in this case, e.g.:
4,14E-10 rs9469779 6 34393845 RPS10 6 34389566 cis A/G G -6,2487055 LLDeep -6,2487055 45 0 172,5 RPS10
Looking at the expression data this should be a possitive effect.

When using the actual most likely genotype calls (by removing the dosage matrix) the direction of the effect sizes are correct again:
8.89536E-13 rs9469779 6 34393845 RPS10 6 34389566 cis A/G G 7.1467109 LLDeep 7.1467109 45 0 0 RPS10 0.8356396 0E0 (0E0) 0.8356396 (0.0837641)0.0

Removing the samples with missing values also resolves the issue.