Uniquely assigned reads seem non-unique

[airichli]

Hi,

The *transcript_counts.tsv file seems to count non-unique reads as unique reads (I use the parameter --transcript_quantification unique_only --gene_quantification unique_only, so only unique reads would be considered). Here I show one example (HDHD5 gene in human):

All reads shown in the figure are considered unique reads from IsoQuant, regardless of v3.5.0 or v3.6.3. However, all these shown reads are mapped to multiple isoforms.

I also include this example as a test data for your check, you can easily reproduce the result in few seconds. The possible reason IsoQuant treats these reads as unique may because the polyA position has one closest match, but all matches just differ in about 10 bp, using polyA position to decide uniqueness may not be appropriate. Could you please explain why IsoQuant treat these reads as unique?

test_data.tar.gz

Looking forward to hearing from you. Thanks.