More unclassified reads in demultiplexing.

[rajwanir]

Hello,

I am trying to basecall some multiplexed r9.4 flongle runs with bonito standard model (dna_r9.4.1), however, I see a greater number unclassified reads upon demultiplexing with qcat.

Bonito + qcat:

Adapters detected in 47391 of 77425 reads
NBD103/NBD104 47391: | ############ | 61.21 %
none 27814 : | ####### | 35.92 %

Guppy + qcat

Adapters detected in 66724 of 77425 reads
NBD103/NBD104 66724: | ################# | 86.18 %
none 10621 : | ## | 13.72 %

The issue seems somewhat related to the one described here but I am using standard model. Could you please suggest on how to resolve this?

Thank you.

[iiSeymour]

Hey @rajwanir

This is most likely a result of how the training set is put together with bonito basecaller --save-ctc .... Each training example has to pass a coverage threshold of 90% but thinking about it this is probably too low. I will have to retrain some models to verify this and get back to you.

[rajwanir]

Thanks @iiSeymour for your prompt response.

[iiSeymour]

Demultiplexing performance should be on par with guppy in v0.3.6.