Duplicated reads (IDs) from nanopore sequencing

Question

[njornet]

Hello,

I basecall the pod5 files generated by the sequencer with dorado and align with minimap2 and in the BAM file I get the exact same read with the same ID and all more than once. Is this normal? If it is, I suppose I have to remove the duplicates right?

Thank you!

[GenoMax]

I get the exact same read with the same ID and all more than once

If you have short(er) reads you could be seeing secondary alignments. Which would be a normal thing. Check the SAM tags

I don't think I have seen what @marco.barr is referring to below.

[marco.barr]

I had the same problem in this iusse https://github.com/nanoporetech/dorado/issues/603 and I solved it as they say here

[GenoMax]

Thanks for the link. We never use "pod5_fail" folders when re-basecalling so we did not see this issue. Current version of MinKNOW now makes a single pod5 folder.

Since original question is about alignments following minimap2 showing duplicate ID, it may likely be due to secondary alignments.

[marco.barr]

it's also true that now I only have one pod5 folder, thanks anyway for clarifying the focus of the issue. Yes, they are probably secondary alignments

[njornet]

I think this is the answer but I can't find in the tags info about primary and secondary alignments. I've only found in the flag that the second entry for each repeated ID corresponds to a supplementary alignment, but from what I've seen that's not the same a what you are saying.

[marco.barr]

How to remove duplicate reads after ONT basecalling from fastq.gz files you could try this

[GenoMax]

If you only have supplementary alignments then they may be caused by reasons mentioned here --> What are Supplementary Reads? I did not mention them originally since they are not main reason for dup id's.

If the alignments are valid you can't simply ignore them as being from duplicate ID's.

[njornet]

I've looked into it and one of the reads with a supplementary alignment has the representative alignment aligning to the + strand and being 1207 bp long and the supplementary aligning to the - strand being 130 bp, but they are overlapping so this is not the indel thing I think

Supplementary Alignments
chr4:433,255-434,368 (-) = 1,115bp @MAPQ 60 NM132 (representative)
chr4:434,228-434,367 (+) = 140bp @MAPQ 33 NM20 (supplemantary)

Here is a screenshot of both reads in IGV, the representative in red and supplemetary in blue.

I've found this thread but they are talking about bwa and I don't think they got anywhere.

Answer 1

[marco.barr]

Hi, I'm also working with ONT data and with dorado. From experience I tell you that some versions of MInKNOW have bad management of fast5 files, wrote duplicate skip folders as both pod5 and fast5. Just filter on unique reads id:s and it should be fine. However I also solved it using updated versions of dorado and MInKnow. They fixed this problem with new versions of dorado and MInKnow. What versions are you using?