WARNING: no modified appropriately-coded values found for ['A', 'C'] in the first 100 reads.

[Gizemmaltun]

Hello, I am trying to use the package. But i am encountering with the below warning message:

WARNING: no modified appropriately-coded values found for ['A', 'C'] in the first 100 reads.
Do you expect this file to contain these modifications? parse_bam is looking for ['A,0', 'CG,0'] but for ['A', 'C'] found only found [].

Consider passing only the motifs and mod codes (e.g. m,h,a) that you expect to be present in your file.
You can use modkit adjust-mods --convert [OPTIONS] <IN_BAM> <OUT_BAM> to update or consolidate mod codes.
See https://github.com/nanoporetech/modkit/blob/master/book/src/advanced_usage.md

No specified number of cores requested. 24 available on machine, allocating all.
Modification threshold of 190 assumed to be for range 0-255. 190/255=0.7450980392156863 will be sent to modkit.
| Step 1: Identify motif locations in GRCh38.d1.vd1.fa 0% | 00:00
| Step 2: Parse regions in 9766b86d69012baef2d90f52445d8fab941dcaca_SQK-NBD114-96_barcode02.bam 0%…
| 0%
The code executed successfully.

I am not sure how to proceed with this. Thank you!

[OberonDixon]

What is the pipeline you are using to make your .bam files? This check is run before we call the modkit pileup or extract modules, and is making sure there are identifiable modifications in the bam tags. I am planning some adjustments to this functionality soon because it does sometimes seem to falsely trigger, but depending on your upstream pipeline this may indicate an issue with the format of the bam files coming in.

If after parsing you can successfully get nonzero values from load_processed or plot functions, you can most likely ignore the warning for the time being.

[Gizemmaltun]

I used dorado basecaller hac,5mCG_5hmCG,6mA. But i have barcoded samples. After basecalling, I did demultiplexing with dorado demux with following command

"dorado demux
--kit-name SQK-NBD114-96
--output-dir ${DEMUX_OUTPUT_DIR}
${INPUT_BAM} --no-trim"

But i lost the alignment. Then i added, --emit-fastq then re-align. Probably in this step i lost the modification. Do you have any suggestions how to demultiplex? Thanks a lot!

[OberonDixon]

@thekugelmeister will be able to help out with this; we've got a working pipeline. I do think that converting back to fastq probably deleted your mod info, I've seen that in the past in other contexts.

[thekugelmeister]

Yes, I can hopefully provide some context! I will note that we are not the developers of dorado; I am happy to provide some basic support and guidance, but I will point you to the dorado team for further assistance.

Loss of alignment

It's interesting to me that you lost alignment even when specifying --no-trim in the demux step. As stated in the dorado trim documentation:

The output of dorado trim will always be unaligned records, regardless of whether the input is aligned/sorted or not.

I would have assumed that this was the root cause of the loss of alignment information, but maybe that was not the case for you?

Loss of methylation information

I have never tried using the --emit-fastq option for dorado demux, but as Oberon stated, that is the most likely place where the modification information disappeared.

How I run dorado

My dorado pipeline is optimized to allow saving some important intermediate files, to avoid unnecessary re-analysis in the future. This may not fit your exact usecase. However, the basic outline might give you a structure that works.

1. Basecall with dorado
   • appropriate model; HAC/SUP, necessary base mods, etc
   • Set modified bases threshold to 0
   • Classify barcodes; don’t trim, don’t demultiplex
   • Output: single unaligned bam file documenting every single modifiable motif in every read, with barcode labels (SAVED FOR LATER)
2. Rest of processing for actual use
   1. Trim with dorado
      • Output: single unaligned bam file documenting every single modifiable motif in every read, with barcode labels, ends trimmed
   2. Align with dorado / minimap2
      • Output: single aligned bam file documenting every single modifiable motif in every read, with barcode labels, ends trimmed
   3. Demux with dorado, no classification
      • Output: multiple demultiplexed aligned bam file documenting every single modifiable motif in every read, with barcode labels, ends trimmed

[Gizemmaltun]

Thanks a lot for your clear explanation! I do have, however, one question. How did you perform the alignment? Because the output from dorado trim will be a BAM file which then needs to be aligned. But dorado/minimap2 requires fastq as an input? Or there is a way to align the bam file as well? Thanks a lot!

[thekugelmeister]

Dorado can perform all of the steps listed above, including alignment. Refer to the dorado aligner documentation for details.