Entering edit mode
5.5 years ago
namkungho87
▴
30
Hi,
Using MatrixEQTL R package, I am performing both cis and trans eQTL analyses in order to find some SNPs associated with two genes of interest (for which we have profiled two specific genes of interest using qRT-PCR). As a result, interestingly, whereas there were NO significant cis-eQTLs for both genes, there were many trans-eQTLs of significance (FDR<0.05).
I tried to look into the detailed data distribution, and realized the distribution is like the below attached picture. Given that there were only two subjects (among 94 subjects) that have ALT/ALT, I am wondering if the result from this analysis is still valid or not...
Please advise me.
Thanks!
I think you spotted the problem. In my experience, MatrixEQTL has the only major drawback that is very sensitive to rare genotypes. I would be very suspicious of those results. In our common practice, we require that the rarest genotypic class has at least five subjects, to consider a result. The threshold is somewhat arbitrary, but it seemed to be working fine for us, and we have more or less the same number of samples as you have (I think 92 or 96, I don't remember right now). It would be nice to hear the opinion of the author of MatrixEQTL about this. I am sure he could provide more insights on this.
Thanks Fabio! Yes, it was exactly what I suspected... I have just asked the author of MatrixEQTL and hopefully will be getting his advice. Thanks anyway!
Please, if you get some nice advice, feel free to post it (also as an answer to your own question), so that everybody (me in the first place!) can read it!
Hi, Fabio. i also encountered the same problem and checked the sample distribution of each genotype (AA, AB and BB). But, my sample size is very small (~16). so, do you think it's appropriate to set the threshold of the rarest genotypic class lower (for example, 3)? and is it necessary the three genotypes(AA, AB and BB) appearance at the same time?
Any comments would be appreciated.