High proportion of noninformative reads in isoquant quantification with nuclear RNA-seq samples

[dawangran]

Hi isoquant developers,
I'm encountering an issue with isoquant quantification where I'm seeing a very high proportion of noninformative reads when analyzing nuclear RNA-seq samples. Here are my read assignment statistics:

2025-11-07 23:50:23,481 - INFO - Read assignment statistics
2025-11-07 23:50:23,481 - INFO - ambiguous: 2,003,938
2025-11-07 23:50:23,482 - INFO - inconsistent: 3,901,428
2025-11-07 23:50:23,482 - INFO - inconsistent_ambiguous: 4,694,691
2025-11-07 23:50:23,482 - INFO - inconsistent_non_intronic: 2,923,910
2025-11-07 23:50:23,482 - INFO - noninformative: 32,654,986 (~71% of total reads)
2025-11-07 23:50:23,482 - INFO - unique: 3,424,641
2025-11-07 23:50:23,482 - INFO - unique_minor_difference: 234,197

As you can see, approximately 71% of my reads are classified as noninformative, which seems unusually high. I'm working with nuclear RNA-seq samples (not whole cell or cytoplasmic RNA).

My questions are:
What are the common causes for such high noninformative read proportions?
Are there any known issues with nuclear RNA-seq samples that could explain this pattern?
Could this be related to the high intronic content typically found in nuclear RNA?
Are there specific parameters I should adjust when working with nuclear samples?

[andrewprzh]

Dear @dawangran

Yes, currently IsoQuant looks primarily for spliced reads, so reads mapping to intronic regions are marked as non-informative. What is the most common assignment subtype that you observe?

As the demand for this kind analysis grows, I will try to change the default strategy and report at least the ID of the assigned gene for such reads in the next release.

Best
Andrey