I have RNA-seq data on a haplotype-resolved cell line and I want to detect genes where there are more reads expressed from one chromosome allele than the other. First I counted informative reads, and then used the binomial distribution, analogous to a coin toss, to determine a p-value. The problem is that highly expressed genes always lead to a low p-value. Is there any way to add a parameter to the binomial distribution to make the tails "wider"? Or use a different distribution?