1
$\begingroup$

I'm trying to analyze data from a genotype-phenotype association study. The genotype data is stored in csv files that record for each variant site and each subject, the alleles that the subject has at that site. A quick online search reveals that 0 means the reference allele, 1 means the most common alternate allele, 2 means the 2nd most common alternate allele, etc.

But I noticed that at some variant sites, most (>90%) subjects have 1/1, i.e. two copies of the alternate allele. I asked around and some suggested I just swap the roles of reference and alternate alleles whenever I find that the alternate allele frequency > 0.5. My questions are:

  1. Can there be a good reason to classify a less common allele as the reference allele? I thought that the reference allele was defined as the most common one in the population.
  2. Is swapping the roles of reference and alternate alleles whenever I find that the alternate allele frequency > 0.5 a standard practice?

Thanks

$\endgroup$
1
$\begingroup$
  1. You are mixing up a bit the terminology here. One thing is a reference allele and another is a major allele. The first refers to the genotype present at a given position in the "reference" genome (e.g. if you work with human data this can be the reference version hg19 or hg38, etc). The second refers to the most common allele in a population. This definition can lead to problems since what is the most common allele can differ among different populations. In this respect, it is perfectly fine to have a reference allele with a lower frequency than another alternate allele.

  2. No for the reasons above. If you mean whether swapping major to minor is standard practice whenever you find out the major allele is actually minor as per the definition (less frequent), then I guess you can but this will lead to inconsistencies whenever you try to merge your data with other data sets, or other people merge their data with your data. Since this for sure happens and issues about strandness are well known, one has to make sure that you make explicit how you treat your data once you release it, and also one has to make sure that major/minor are defined the same way when merging data, and do the fix whenever needed.

$\endgroup$
1
$\begingroup$
  1. Yes, depending on how the reference allele was determined. If, for example, you want to refer to hg38 as the reference genome, then it's important to use the hg38 allele as reference.

  2. Without stating the reference allele convention? Not that I'm aware of. However many people refer to minor/major alleles rather than non-reference alleles, in which case it is assumed that the major allele is the most common allele.

As long as you state what your reference genome and/or population is, then it's fine to choose your own convention about reference alleles. But please be consistent and explicit about what you've done. One of my biggest struggles in GWAS has been working out the correct strand orientation for stated variants (important when merging datasets), and it doesn't help for complementary variants (i.e. A/T, C/G) when the ref/alt alleles are arbitrarily chosen.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.