False variant calls due to alignment (Allele specific expression, aka ASE)

[Sergio.pv]

Dear Community
I am trying to identify genes having allele specific expression from RNAseq data using GATK's ASEReadCaller and MBASED (https://genomebiology.biomedcentral....059-014-0405-3).

ASEReadCaller uses BAM files to produce a table with rows for SNP sites and columns for the alternative and reference allele counts.

If the alternate allele has at least 5 counts, a variant is called (this is an arbitrary threshold; some studies use 3 counts).

Heterozygosity is defined as sites with a minimun of 10 total counts and min 5 counts per allele.

To apply MBASED it is suggested to remove heterozygous sites being too close to each other (within 10bp), as this is evidence of false variant calls due to alignment.

I am having trouble understanding this concept.

The way I see it, a false variant call results from reads with a mismatch being correctly aligned to the reference genome; that mismatch will be considered as the alternate allele if it matches to the actual non reference allele and then a variant is called (according to a given threshold). Now, if a read carries more than one mismatch, then it will potentially produce more than one false variant call.

Since reads are not too long and there is a given RNAseq error rate, the easiest explanation for variant calls that are too close from each other is that reads mapping to that location contain mismatches.

I would greatly appreciate if someone can tell me if I am reasoning through this correctly

Thanks in advance!