Trouble with gapped alignment about hisat2

[jdidion]

After setting the minimum score low enough, I finally get an alignment:

SRR3106538.397648993 83 chrM 303 255 57S8M1I85M = 15 -438 GTCTGCACAGCCGCTTTCCACACAGACATCATAACAAAAAATTTCCCCCCAACCCCCCCCCCCCTCCCCCCGCTTCTGGCCACAGCACTTAAACACATCTCTGCCAAACCCCAAAAACAAAGAACCCTAACACCAGCCTAACCAGATTTCA +++$+?$?$???+++?++?$++?+++?+++++?+++??????????+??+?+??????????????????????????????????????????????????????????????????????????????????????????????????? AS:i:-58 ZS:i:-66 XN:i:0 XM:i:0 XO:i:1 XG:i:1 NM:i:1 MD:Z:93 YS:i:-6 YT:Z:CP NH:i:1
SRR3106538.397648993 163 chrM 15 255 151M = 303 438 CACCCTATTAACCACTCACGGGAGCTCTCCATGCATTTGGTATTTTCGTCTGGGGGGTATGCACGCGATAGCATTGCGAGACGCTGGAGCCGGAGCACCCTATGTCGCAGTATCTGTCTTTGATTCCTGCCCCATCCTATTATTTATCGCA ??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????+????+???++?+???+???++?+?++++???# AS:i:-6 XN:i:0 XM:i:1 XO:i:0 XG:i:0 NM:i:1 MD:Z:131T19 YS:i:-58 YT:Z:CP NH:i:1

However, strangely, HISAT2 is choosing to soft-clip 57 bp so as to only open a 1 bp gap, rather than preferring the better overall alignment with a 6 bp gap. For reference, here is BWA MEM's alignment:

SRR3106538.397648993 163 chrM 15 60 151M = 251 381 CACCCTATTAACCACTCACGGGAGCTCTCCATGCATTTGGTATTTTCGTCTGGGGGGTATGCACGCGATAGCATTGCGAGACGCTGGAGCCGGAGCACCCTATGTCGCAGTATCTGTCTTTGATTCCTGCCCCATCCTATTATTTATCGCA ??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????+????+???++?+???+???++?+?++++???# NM:i:1 MD:Z:131T19 AS:i:146 XS:i:83
SRR3106538.397648993 83 chrM 251 60 59M6I86M = 15 -381 GTCTGCACAGCCGCTTTCCACACAGACATCATAACAAAAAATTTCCCCCCAACCCCCCCCCCCCTCCCCCCGCTTCTGGCCACAGCACTTAAACACATCTCTGCCAAACCCCAAAAACAAAGAACCCTAACACCAGCCTAACCAGATTTCA +++$+?$?$???+++?++?$++?+++?+++++?+++??????????+??+?+??????????????????????????????????????????????????????????????????????????????????????????????????? NM:i:10 MD:Z:12A33A2A9T85 AS:i:113 XS:i:45

Maybe this is due to the fact that BWA MEM performs local realignment? Is that something HISAT2 does, or that is planned for the future?

[infphilo]

Thank you for reporting this issue, and sorry for taking this long time to get back to you (I've been swamped). HISAT2 previously allowed insertions and deletions up to 3 bps long. I changed it to allow indels of arbitrary length as long as the alignment score is above or equal to the minimum alignment score. The fix is available in the master branch if you'd like to give it a try. I'll try to release a new version of HISAT2 in a couple of weeks.

For your read, the new implementation has the following alignment:

0 0 MT 251 255 52M5I8M1I85M * 0 0 GTCTGCACAGCCGCTTTCCACACAGACATCATAACAAAAAATTTCCCCCCAACCCCCCCCCCCCTCCCCCCGCTTCTGGCCACAGCACTTAAACACATCTCTGCCAAACCCCAAAAACAAAGAACCCTAACACCAGCCTAACCAGATTTCA IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII AS:i:-46 XN:i:0 XM:i:3 XO:i:2 XG:i:6 NM:i:9 MD:Z:12A33A2A95 YT:Z:UU NH:i:1