Have been thinking about this for a while. As you mentioned, now added. Option -y.

Closing via 08bd212.

Thanks for the quick response, Heng!

Here's a more involved idea… do you think minimap2 could chain through alignments of reads in the same barcode, to create an alignment of the full molecule? For CIGAR alignments, perhaps N could be used to indicate the gaps between reads.

That would amount to a 10x mapper, which is distinct from typical short-read mapping. The complication lies in that different genomic regions may be associated with the same barcode. You have to identify these "read cloud" first and then do mapping. Reads in these cloud have no specific order, either.

In general, the 10x mapper has room for improvement. One particular problem is that it uses bwa-mem, which is not good at all mappings. Minimap2 has the potential, but it will take significant efforts to implement. I think someone from the Bonnie Berger group has a better mapper. I thought it was on bioRxiv but could not find it.

http://cb.csail.mit.edu/cb/ema/

Thanks, @mcshane. This is what I was talking about. The first author gave a talk at Broad. I think it is quite good, though I haven't used it personally.

Yeah, I haven't tried yet either, but it does look interesting.

Thanks for the pointer to EMerAld (EMA). I'm not familiar with this aligner. I'll check it out.

Back to minimap2 -y. Is the FASTQ comment copied to the SAM tag even when it's not formatted in a valid SAM tag format? In particular, when the comment is in Illumina bcl2fastq format (1:N:0:ATCACG), it'd be helpful not to copy it to the output and create an invalid SAM file.

I'm writing a pipeline where the FASTQ files may either be in bcl2fastq format or Longranger basic format, and it'd a nice bonus for the pipeline to be able to use minimap2 -y with both, and not have to detect which format the reads are in.

I am afraid that you have to do this on your side.

Okay. Thanks, Heng. If I were to implement this feature, would you accept a pull request?

Thanks. I will surely consider. Please don't introduce new dependencies, though.