I am using snpeff for variant annotation and effect prediction. For my genes of interest, I am looking for the disrupted gene models in the population (snp data for 50 individuals). Loss of function analysis or gene disruption is done based on SNPeff criteria's- loss of start, stop gain etc. Few of the positions are heterozygous. My question is how should I treat these positions? Suppose for a heterozygous position C/G, G causes an early stop codon. We don't know which allele is expressed or dominating? So for these cases, I am concluding that the gene model is broken. Is this correct to say?