--score function in PLINK2

[Sander W. van der Laan]

Hi,

Just a quick question regarding the --score function described here. 

Given that the user is providing alleles, is PLINK considering the orientation/flipping of alleles in the process of calculating the score? 

If not: are you considering adding this (as an option)? 

Thanks

Sander

[Christopher Chang]

Can you clarify exactly what you mean by this, with a worked example?

[Sander W. van der Laan]

Sure, I don't have a worked example, other than a description.

Suppose I have this PGS:

SNP EffectAllele Effect

SNP1 A 0.34

SNP2 T 0.02

SNP3 C -0.98

SNP4 A 0.56

SNP5 G 0.06

And I have these genetic data (just showing example frequency table below), and suppose the B-allele is the 'coded' allele, in other words the dosages/probabilities are relative to the B-allele.

SNP A-allele B-allele B-alleleFreq

SNP1 A C 0.51

SNP2 T A 0.46

SNP3 T C 0.09

SNP4 T A 0.35

SNP5 T G 0.23

For SNP1, EffectAllele=A with Effect=0.34, but in my data SNP1 has A-allele=A, and B-allele=C. So my data is coded according to the B-allele. Is --score aware of this? So is it going to take the dosages of the non-coded allele in that case? 

Thanks

Sander 

[Sander W. van der Laan]

Not sure if I am making myself clear... :-)

[Christopher Chang]

--score results aren't going to change based on which allele is coded A1/A2.

[Sander W. van der Laan]

Thanks for the clarification. 

[Sander W. van der Laan]

Follow-up question. I am getting the error below: 

```

Error: --score variance-standardize failure for ID 'rs61886614': estimated
allele frequency is zero, but not all dosages are zero. (This is possible when
e.g. allele frequencies are estimated from founders, but the allele is only
observed in nonfounders.)

```

This causes the calculation to stop. How do I tackle these types of errors? Basically, I don't get scores for chromosomes 10, 11, 12, because of this error. Should I remove all the variants where frequency = 0?

[Christopher Chang]

(oops, somehow missed this when it was initially posted)

variance-standardize has unstable behavior with very-low-MAF variants; 0 MAF is just an extreme case of this.  If an allele-frequency filter wasn't previously applied to your dataset, yes, you should include one in this operation.