Gestion of microexons by bambu

[wdesaintjean]

Hello,
I want to know how bambu manage the microexons in the generation of the transcriptome.
I tested bambu with:
- direct minimap2 alignment
- minimap2 alignment corrected by the MisER software (https://github.com/zhenLiuXplr/MisER-project/tree/v1.0.0) to realign the error mapping of minimap2 on microexons

But the results are pretty similar: bambu seem to already detect and quantify annotated isoforms containing microexons, even if the microexon is not aligned by minimap2 but only annotated.

I didn't find any explanation in the paper or the README, so I dare to ask you some supplementary explanation.

Best regards,
William

[lingminhao]

Hello @wdesaintjean, this is controlled by a parameter in bambu called min.exonDistance (default 35bp). As long as the microexon is < 35bp, bambu treats the read as similar enough to the annotated isoform and still quantifies it onto that transcript, even if the microexon isn't present in the alignment. This is also why your minimap2-only and minimap2+MisER results came out comparable. You can adjust the threshold via opt.discovery = list(min.exonDistance = ...) if you want stricter matching.

Let me know if it answers your question.
Hope it helps!
Min Hao

[wdesaintjean]

Hello @lingminhao,
thank you very much for your answer, it's exactly what I wanted to know !!!

Just to precise, in the unknown isoforms find by bambu, we can see the differences before and after the realignment on microexons.

Thank you again, and have a good day