Is it ok to use genotype replicates for association studies?
0
0
Entering edit mode
2.4 years ago
rimgubaev ▴ 270

Hello everyone!

I got 90 individuals genotyped in 3 replicates (genotyping by sequencing), so as a result, my vcf file contains information on 270 samples. The phenotype data were collected for 90 individuals. Then I try to run the mixed linear model in TASSEL and each of 3 genotypes replicates got the same phenotype data. As I understand this is not ok since the phenotype variance getting smaller artificially.

And my problem solution now looks as follows:

  • I make the permutations, changing the phenotype data for each of 3 genotype replicates randomly for 500 times and running the MLM to find significant associations on random combinations.
  • Then I calculate p-value as follows: I divide the number of iterations with the number of significant SNPs equal to or greater than the number of significant SNPs obtained with using original phenotype-genotype combination by the number of iterations.

I obtain a pretty small p-value, around 0.02 and then also make a binomial test in order to ensure that it is unexpectable to obtain that number of significant SNPs by random chance.

If you ever had a similar situation please share some advice. I also would be grateful if you share your opinion.

Thank you.

GWAS permutation test TASSEL • 471 views
ADD COMMENT
1
Entering edit mode

Please format your post better so it is easier to read - split into paragraphs with breaks where there is a slight change in context. When the content is a "wall of text", it discourages people from reading and by extension, getting to the core of your problem.

I've tried to format your post as best as I can, but please feel free to edit it and make any changes you think would be appopriate.

ADD REPLY
0
Entering edit mode

That sounds not appropriate, and I think you should merge those replicates to keep one record per individual rather than 3. Are the genotypes between replicates mostly concordant?

ADD REPLY
0
Entering edit mode

Yes, most of them (>75%) cluster together but others do not (I find out it by building an NJ tree). It s a plant sample so each replicate corresponds to a plant from a certain line, so it is biological replicate not technical. I tried to merge genotypes but in that case, I lose a huge amount of meaningful SNPs.

ADD REPLY

Login before adding your answer.

Traffic: 2982 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6