Redundant isoforms

[KMbio43]

Dear Andrewprzh,

Thank you very much for developing such a useful software. I have a question I'd like to ask you.
For example, if two transcripts have exactly the same intron structure but differ only at the 5' or 3' end, under what circumstances would IsoQuant merge these two transcripts into one?
For example:

1. If the 5' ends of the two transcripts are within 30 bp of each other, and the 3' ends are also within 30 bp, I believe they should be merged into one transcript.
2. If the 5' ends of the two transcripts are within 30 bp of each other, but the 3' ends are far apart, I don't think they should be merged.
3. If the 3' ends of the two transcripts are within 30 bp of each other, but the 5' ends are far apart, I also don't think they should be merged.
4. If both the 5' and 3' ends of the two transcripts are far apart, despite having identical intron structures, they should not be merged.
   These are the scenarios I am considering. Does IsoQuant have any parameters that control this behavior? I have not found any related settings in the help documentation.

Best regards,
Kai

[andrewprzh]

Dear @KMbio43

Sorry for the delayed response.

At the moment IsoQuant still implements a rather simple conservative strategy in terms of detecting transcript ends and this is one of the direction we are currently working on. In general, I agree with the strategy you wrote, but the main problem is correctly identifying TSS and polyA coordinates. The main challenges here are: (1) reads can be truncated and (2) polyA site positions have quite some variability, especially in ONT reads. Thus, reporting isoforms with identical TSS/polyA sites can lead to a substantial increase in false positives.

In the current version, if a reference annotation is provided and an a predicted isoform has an intron chain perfectly matching one of a known isoforms, the known isoform will be reported "as is" and no alternative TSS/polyA sites will detected.

For novel isoforms having identical intron chain, IsoQuant will attempt to detect alternative polyA sites if they are more than 50 bp away. However, detecting alternative TSS positions is not done due to read truncation. Thus, in de novo more (without reference annotation), polyA sites can be detected even more accurately.

In summary, at the moment IsoQuant focuses on alternative splicing rather than polyA sites, but we hope to update these algorithms in the nearest future.

Best
Andrey

[KMbio43]

Thank you very much for your reply.
Because there are already some full-field sequencing methods that protect the 5' end of RNA. For example, capTrap-seq and TeloPrime-based protocol with cDNA Amplification Kit v2 both operate with a 5' cap. Therefore, the RNA produced by this sequencing method is generally an intact RNA at both ends. Therefore, I hope you will consider this situation in future releases. In addition, as for splicing junctions, I believe that when sequencing is accurate enough, two intron cuts should not be classified as one intron, even if they are within a few bases of each other, because even a base difference in the CDS region may result in frameshift mutations. I don't know if my understanding is correct, I hope to discuss it with you.

[andrewprzh]

Yes, we also plan to incorporate CapTrap data and similar protocols which ease TSS detection. Unfortunately, due to the amount ongoing projects it's a bit hard to predict when these improvements will be implemented.

As for splice junctions, it mostly comes to the question how accurate the aligner is. We did some analysis of PacBio and ONT reads (https://genome.cshlp.org/content/32/4/726.short) and also implemented a splice correction algorithm inside IsoQuant. So I can say IsoQuant detects them rather accurately. However, if you are sure in your data, you can, for example, set --delta 0 so the splice junctions will be derived right from the BAM file without any slack.

Feel free to continue discussion here or via email (can be found in my profile).

Best
Andrey

[KMbio43]

Thank you very much for your reply. I look forward to your updates to the software, and if I have any further questions, I hope you will be happy to help me.

Best
Kai