Minimap2: "Number of residue matches" is off by 2x without cigar string

[kpalin]

Possibly documentation issue but hopefully some sort of fix needed: The 10th column on paf output is off by 2x if no aligment is requested. I know paf spec let's those be "approximate" but I'd rather have them way off (like zero or minus one) if they're this much wrong.

See my_read.fastq.zip for difference between 48% identity and 88% identity.

$ minimap2 /data/reference-genomes/lambda_phage/lambda_ref.fasta my_read.fastq |cut -f 10-11
20319 42448

$ minimap2 -c /data/reference-genomes/lambda_phage/lambda_ref.fasta my_read.fastq |cut -f 10-11
37736 43060

[lh3]

That is a documentation issue. I will think about how to improve.

Without base-level alignment, column 10 gives the number of matching bases in the chain. It is often far off. You should look at the dv:f tag instead. It estimates per-base sequence divergence more accurately.

[lh3]

I have added to miniasm/PAF.md that col10 and col11 can be "highly inaccurate". There is little I can do here unfortunately.

[SamStudio8]

Am I right in thinking that (1 - dv:f) * 100 would give an approximate identity?

[lh3]

Yes, approximate.