Suggestion for Adjusting Scoring Logic in long_read_assigner.py (extra_left and extra_right)

[biosalt-cc]

Hi, IsoQuant is an incredible tool that has significantly streamlined long-read transcriptome analysis.

Issue Description:
While using IsoQuant, I noticed a potential improvement in the logic for calculating extra_left and extra_right in src/long_read_assigner.py.

File and Lines:
src/long_read_assigner.py, lines 187–188

Original Code:

extra_left = 1 if read_region[0] + self.params.delta < transcript_start else 0
extra_right = 1 if read_region[0] - self.params.delta > transcript_end else 0

Suggested Update:

extra_left = 0.5 if read_region[0] + self.params.delta < transcript_start else 0
extra_right = 0.5 if read_region[1] - self.params.delta > transcript_end else 0

Please let me know if this suggestion aligns with the intended design.

[andrewprzh]

Hi @biosalt-cc

Thank you for the feedback!

I am really amazed by the fact you were able to dig this deep. In general, I agree that alternative splicing events are not the same weight as terminal exon elongation. Do you have any specific cases where this comes into effect?

Best
Andrey

[biosalt-cc]

Hi, @andrewprzh
Sorry, I didn't make the question clear.

I think [extra_right] is intended to assess whether the end of the read region extends beyond the end of the transcript region. However, it incorrectly uses [read_region[0]] (start position) instead of [read_region[1]] (end position).
This mistake may result in an inaccurate calculation of inconsistencies on the right side of the read, potentially leading to incorrect isoform selection.

In fact, while testing, I found that some reads were not correctly assigned to ISOforms, and I tracked down this potential problem step by step.

However, please double check if my understanding is correct.

[andrewprzh]

Dear @biosalt-cc

Yes, you are right, that's a bug.
I will fix it ASAP and will try to make a new release until the end of the year.

Best
Andrey