Repository navigation
High NDR values when quantifying transposable elements in exosomes #536
Description
Dear Bambu team,
Thank you for making such an amazing tool available!
I'm analyzing ONT-cDNA data of porcine exosomes with Bambu for downstream transposable element (TE) quantification, and consistently get a high recommended NDR (~0.685) with both the pre-trained human model and sample-specific training.
Could you please advise:
- Should I set NDR = 1 and min.txScore.singleExon = 0 for downstream TE overlap analysis (e.g., with RepeatMasker)?
- Would you recommend sticking with the human model in this case?
I would greatly appreciate your thoughts and suggestions.
Best regards,
Yu
Bambu warning for sample specific training:
Bambu warning for pre-trained human model:
Activity
SuiYue-2308 commentedon Apr 8, 2026
Hi,
Sorry for the late reply.
Setting NDR = 1 will report all transcripts, including those with low confidence, which I generally don’t recommend. However, for transposable elements (which often include single-exon transcripts), I suggest using min.txScore.singleExon = 0. You can then use the resulting txScore to manually evaluate their confidence.
Not sure about the differences between human and porcine annotations: I recommend using a porcine closely related annotation to train the model before inputting it into Bambu for the best results.
Best regards,
Yue