PAF format and identity problem

[jianshu93]

Hi team,

mashmap -k 16 --perc_identity 85 -r GCF_018619585.1_ASM1861958v1_genomic.fna -q query_test5.fasta

minimap2 -x map-hifi -k 16 -w 21 -o try.paf GCF_018619585.1_ASM1861958v1_genomic.fna query_test5.fasta

mashmap3 PAF format is strange:

[jzhao399@atl1-1-02-017-36-2 test_superani_minimap2]$ cat try.paf
test05 2880 3 2863 + NZ_JAHHEG010000008.1 2669591 28003 30863 2667 2860 60 tp:A:P cm:i:277 s1:i:2667 s2:i:0 dv:f:0.0002 rl:i:0
[jzhao399@atl1-1-02-017-36-2 test_superani_minimap2]$ cat mashmap.out
test05 2880 0 2880 + NZ_JAHHEG010000008.1 2669591 26319 29199 92 2880 19 id:f:0.988335 kc:f:1.06012

First is the minimap2 PAF, Second Iis mashmap3 PAF. Same reference and query, column 10 equals the number of sequence matches, and column 11 equals the total number of sequence matches, mismatches, insertions and deletions in the alignment, according to the page here: https://github.com/lh3/miniasm/blob/master/PAF.md, I have 92 for mashamap3, 2667 for minimap2. I understand those are differnce algorithm, but difference by this much? Identity by minimap2 is coulmn 10/11, which is 93.2% but mashmap3 report 98.8%. Any idea why?

Thanks,
Jianshu

[bkille]

Hi @jianshu93!

MashMap doesn't perform base-level alignment, so we don't report the number of matched bases. And as the identity predicted by MashMap is gap-compressed identity, it wouldn't be correct to report the number of matched bases as the product of column 11 and the predicted identity.

Column 10 contains the numerator of the minhash formula for the first map in a chain of segment mappings. Its not particularly informative for chained mappings, and perhaps we should make it the average over all mapping in a chain, but nevertheless it does not represent matched bases.

Also, it is interesting that in your example, the "blast identity" of the minimap2 mapping is 93.2%, but the "Approximate per-base sequence divergence" is 0.0002.

Finally, I notice that your query sequence is only 2880 bp long. I would suggest making sure that the MashMap segment length -s (default 5kbp) is not larger than your query.