some compare result of minimap2 '-x sr' mode of v2.18 and v2.20

[wangyugui]

some compare result of minimap2 '-x sr' mode of v2.18 and v2.20

genome: GRCh38.p13
data: Illumina PE150

a) performance: big change in runtime.
v2.18
real 213.72(3:33.72), user 12295.66, sys 169.28
v2.20
real 323.90(5:23.90), user 17924.58, sys 225.08
but memory usage is almost same.

b) mapping result: big change in 'properly paired %' and 'singletons %'

v2.18
mapped % 97.16%
mapq>=1 % 92.69%
properly paired % 95.98%
singletons % 0.30%
v2.20
mapped % 96.34%
mapq>=1 % 92.01%
properly paired % 92.63%
singletons % 1.07%

c) misc
in man file, '-r50' is in '-x sr' mode,
but '-r100' is in source(mo->bw = mo->bw_long = 100;).

[lh3]

Thanks. On my dataset, I didn't see this. v2.20 is slightly more accurate on the contrary. Perhaps a difference between datasets. Either way, you may try to add -xsr -e0.

[wangyugui]

in this dataset, '-xsr -e0' help the a) performance, but few changes in the b) mapping result.

v2.20 with -xsr -e0
a)performance:
real 291.60(4:51.60) user 17970.39 sys 224.20
b)mapping result:
mapped % 96.33%
mapq >=1 % 92.01%
properly paired % 92.62%
singletons % 1.07%

There is no '-e0' option in minimap2 v2.18, so no test result of v2.18 with -e0 option.

[lh3]

Could you check the github HEAD to see if this is fixed on your data? Thanks.

[lh3]

PS: if it is ok on your data, I will release v2.21 for this bug fix.

[wangyugui]

Very good result with this fix.

minimap2 2.20 r1069 without -e0
a) performance
real 204.98(3:24.98) user 11504.85 sys 183.21
b) mapping result
mapped % 97.17%
mapq>=1 % 92.70%
properly paired % 96.00%
singletons % 0.29%

minimap2 2.20 r1069 with -e0
a) performance
real 202.34(3:22.34); user 11614.83; sys 188.57
b) mapping result
mapped % 97.17%
mapq>=1 % 92.69%
properly paired % 95.98%
singletons % 0.30%

By the way, in the coming 2.21 release, please change the man file too?
c) in man file, '-r50' is in '-x sr' mode,
but '-r100' is in source(mo->bw = mo->bw_long = 100;).

Best regards

[jmarshall]

PS: if it is ok on your data, I will release v2.21 for this bug fix.

@lh3: I am about to send you a (more complete) fix for #573 (CIGAR stuff) if you would like to wait a moment to consider adding that. [ETA: this is now PR #777.]

[lh3]

Thanks, @jmarshall. I will wait. I need to run some large-scale tests anyway before each release. It will take some time.