I like using Minimap2 for our Nanopore RNA sequencing a lot. I tried deSALT and and GraphMap but Minimap2 seems to work better for noisy direct RNA-Seq ONT data.

I usually align the reads to both genome and transcriptome reference. However, the transcriptome mapping didn’t behave as I expected. For the ONT RNA-Seq data the Minimap2 manual led me to use:

minimap2 -a -x map-ont -k 12 -p 1 -u f --secondary=yes ref.fa.mmi reads.fastq.gz

The map-ont doesn’t allow to splice the reads, -k 12 helps with the read noise, -p 1 outputs only same scoring alignments (one as primary, rest to secondary mappings) and --secondary=yes allows to output secondary alignments. The highlighted parameter -u f should force forward strand only mapping. Again, from the manual: For Iso-seq, Direct RNA-seq and tranditional full-length cDNAs, it would be desired to apply -u f to force minimap2 to consider the forward transcript strand only. Therefore I was expecting to see forward strand mapping only since I used transcriptome fasta as a reference. However, this was not a case. A simple scan with samtools view -f 16 showed a reasonable number of mappings. What?! I thought it’s some kind of feature of the data and that Minimap2 is doing everything correctly but somehow the reads mapped to the reverse strand do not map anywhere else so it puts them to the reverse strand as a last resort. This all changed with our latest spike-in only (SIRV) experiment.

In this case I knew there should not be any reverse strand mapping whatsoever. But to my surprise, the command above gave me quite a lot of reverse strand mapping. So I dug a bit deeper on where it the problem and I found it. By accident I found a Ubuntu Minimap2 manual (later on I also found the official one here) where it says:

-u CHAR How to find canonical splicing sites GT-AG - f: transcript strand; b: both strands; n: no attempt to match GT-AG [n]

So in reality, -u f has nothing do to with forward strand mapping only. Or at least not in the way I would expect it. It basically looks for predefined splice sites but only in the same direction as is the read. So it makes sense only in genomic spliced alignments, nowhere else. Whoops. The parameter I was looking for turned out to be --for-only which again, by the Ubuntu manual says:

--for-only Only map to the forward strand of the reference sequences. For paired-end reads in the forward-reverse orientation, the first read is mapped to forward strand of the reference and the second read to the reverse stand.

After removing the -u f and replacing it with --for-only I finally got the alignments I was hoping for and absolutely no alignments on the reverse strand. Note: There is also --rev-only option. It should be possible to find this option in the command-line Minimap2 manual if you use:

man ./minimap2.1' for detailed description of these and other advanced command-line options. 

but this never worked for me since I install Minimap2 using Conda. On the other hand, I didn’t expect the brief manual wouldn’t mention it. You will also find hardly anything at the Minimap2 GitHub.

Obviously, this is not an issue of Minimap2 because it is in the manual after all. But, you have to really look for the manual in order to find it which I think it’s not the best. Hopefully this will help you to figure out why you are not getting your expected alignments (and you are the same ignorant as me).