How to remove paralogous alignment from RNAseq data?

Question

[kirannbishwa01]

I am looking for a method to remove paralogous alignment from aligned RNAseq data. I have already applied mapQ filters.

The main way to distinguish the paralogous in genome resequence data is to look for the depth of the coverage and select for the regions that have more than 97.5th percentile or 4 times the SD coverage from the mean coverage. But, this method cannot be applied to RNAseq data.

In contrary I was thinking if the number of alleles at a locus can be used as a method to detect the paralogous reads in RNAseq data. Reason: in a diploid genome at any site there cannot be more than 2 alleles except when there is sequence error. So, if there are more than 2 alleles at a site with equal coverage

1) this could help us identify the potential paralogous alignment site.

2) The next step would then to find the reads that are paralogous vs. non-paralogous.

But, I have found no tools to work this out.

Has, any one had similar problems - solution.

Thanks,

Answer 1

[Fabio Marroni]

The main way to distinguish the paralogous in genome resequence data is to look for the depth of the coverage and select for the regions that have more than 97.5th percentile or 4 times the SD coverage from the mean coverage. But, this method cannot be applied to RNAseq data.

Your statement is only true if the assembly against which you are aligning contains only one of the possible paralogs. If all paralogs have been correctly assembled then you would either a) align each read correctly to each paralog, or (more likely) b) align reads corresponding to one gene to several positions each one corresponding to one paralog. Thus, maybe following multiple alignments patterns you might discover paralogs.

This is just a thought, I have no great experience on that.