demultiplexing with guppy_basecaller and guppy_barcoder using --detect_mid_strand_adapter --detect_mid_strand_barcodes produce > 90% of unclassified reads

Question

[andres.firrincieli]

A similar question has been already posted on the ONT community forum but without answers form guppy devs.

I am using guppy 6.0.1

In short, when demultiplexing during basecalling using the options --do_read_splitting --detect_mid_strand_adapter --detect_mid_strand_barcodes

./guppy_basecaller -c dna_r9.4.1_450bps_sup.cfg -i $SCRATCH/barcode/fast5 -s $SCRATCH/barcode/fastq -x 'auto' --recursive --barcode_kits "EXP-NBD104" --trim_barcodes --trim_adapters --do_read_splitting --detect_mid_strand_adapter --detect_mid_strand_barcodes

90% of the reads are placed within the folder unclassified suggesting that demultiplexing fail if barcodes and adapters are in the middle of a read. In a second run, by omitting --do_read_splitting --detect_mid_strand_adapter --detect_mid_strand_barcodes:

./guppy_basecaller -c dna_r9.4.1_450bps_sup.cfg -i $SCRATCH/barcode/fast5 -s $SCRATCH/barcode/fastq -x 'auto' --recursive --barcode_kits "EXP-NBD104" --trim_barcodes --trim_adapters

only a small percentage of reads is found as unclassified.

Then I decided to run guppy_barcoder on demultiplexed fastq files obtained from the second command line as follows:

./guppy_barcoder -i $SCRATCH/barcode/fastq/pass/barcode01 -s $SCRATCH/barcode/fastq/pass/barcode01/mid_trimmed -x 'auto' --recursive --barcode_kits "EXP-NBD104" --trim_barcodes --trim_adapters --detect_mid_strand_adapter --detect_mid_strand_barcodes

Surprisingly, more than 90% of the reads that were correctly classified as barcode01 are now unclassified, confirming that if the adapter is detected in the middle of the read, guppy_basecaller does not perform any splitting but detect the read as unclassified.

Now, my main questions are:

Is it possible that more than 90% of the reads have an adapter in the middle of the sequence? I know chimeric reads are possible in nanopore sequencing but 90%... this is quite scary.

Does --do_read_splitting works as intended (i.e. Perform read splitting based on mid-strand adapter detection)?

Thank you

[Yuri]

Hi andres.firrincieli did you ever figure out why guppy was resulting in 90% of reads in unclassified folder?

I am running into the same issue, and am curious what could be the cause...

[andres.firrincieli]

Hi Yuri

I still have no idea what could be the cause but switching to guppy v6.2 solved this problem