Entering edit mode
6.6 years ago
hxlei613
▴
100
Hi, I've read many papers about allele specific expression(ASE). Most of them have phasing data so it is reasonable for them to get allele specific expression genes. For those who don't have phasing information, they just perform test on heter-SNPs but they also get corresponding genes without details. Is this ok ? I think they just consider a gene as an ASE gene if all SNPs in that gene show ASE. For example a software called ASEQ https://www.ncbi.nlm.nih.gov/pubmed/25889339, it doesn't need maternal or paternal information but output ASE genes.
I'm very happy for any comments. Thanks!
I gave a quick look to the paper, and it is true that they never mention haplotype phase. However, HapMap data is often phased. So, I don't know if they are implicitly using phased data. You are right. Theoretically, you need the phase to infer ASE. A quick and dirty approximation is to assume that if you see several SNPs in your gene with ASE, the your gene has ASE. Why is this quick and dirty? Because e.g. if you have 4 SNPs, all of them having a 40/60 ratio, if you don't have the haplotype you will usually assume that all the SNPs have ASE in the same direction, while maybe it is not the case. Even if all the SNPs in the gene show ASE, this might be due to random errors. Due to the fact that you do not know the phase, you are assuming that any deviation from 50/50 is due to ASE and you are also probably going to assume that the alleles with the lower expression lay on the same haplotype, while there is no reason to believe this. Depending on the haplotype structure of the reference and of the study sample, this quick and dirty approach may lead to small or large errors in the estimates. Today there are phasing tools to short-range phasing of NGS data (such as hapcut). This would already be better than nothing.