f4a8766283
r362: fixed overestimated chaining score
...
Caused by ilog2_32(0)=-1. This bug was fixed once and reoccurred as I was
tuning the score function but forgot to apply the fix.
2017-09-14 10:15:22 -04:00
c59b0781bc
r280: output introns as "N" in the cdna mode
2017-08-09 11:45:02 -04:00
1a7d782131
r273: cdna mapping mode for testing
...
Differences from the typical mapping mode:
* banded alignment disabled
* log gap cost during chaining
* zero long-gap extension during alignment
* up to 100kb (by default) reference gap
* bad seeding not filtered (to tune later)
2017-08-08 11:31:49 -04:00
fc965805f7
r215: bring back a log gap component
...
Otherwise chaining may more often break a long gap into several gaps.
2017-07-28 00:17:19 -04:00
f2ef48878a
r202: trim bad chain ends before extension
...
This fixes a few more FP long INDELs towards the end of alignments.
2017-07-25 19:53:19 -04:00
21ca564112
r201: fixed a minor chaining issue
...
Chaining looked at the end of a chain, but the end may not be the best. We now
go back to find the max.
2017-07-25 18:26:51 -04:00
215e92ed7b
r200: reduce long gaps in chaining
...
Every seed can initiate a chain.
2017-07-25 17:32:54 -04:00
b530ade333
r199: changed to linear gap cost for chaining
...
The old cost doesn't penalize long gaps enough. Will also drop seeds close to
the edge in the next commit.
2017-07-25 15:35:10 -04:00
eeeb2ffb68
r174: make max-chain-skip work
...
The max-chain-skip heuristics did not work due to a bug. Without this
heuristics, chaining is too slow for long-read overlap.
2017-07-12 10:08:06 -04:00
7598809577
r171: reduced log gap cost at chaining
...
The cost is so large that it discards too many valid seeds without HPC k-mers.
This change may introduce false long gaps to reference mapping. We have another
mechanism mm_filter_bad_seeds() to protect against this. In addition, minimap2
is not that bad to have long gaps. Some other aligners are worse.
Still need tuning in future.
2017-07-11 14:57:49 -04:00
782449975d
r168: fixed a bug in long join: a[] not sorted
...
Also added length requirement for long join and changed -g in the ava mode
2017-07-09 12:14:20 -04:00
3f6a0b0b5c
r162: improved chaining accuracy
2017-07-08 14:29:36 -04:00
2338e887d9
finished the first draft of manpage
2017-07-01 11:25:54 -04:00
d274e1b743
backup
2017-06-29 12:58:52 -04:00
c8d122bcdb
backup
2017-06-29 11:11:15 -04:00
a25866c25c
r96: min_cnt still wrong in chaining
2017-06-28 11:03:03 -04:00
bf0e8199e2
r94: min_cnt is tested in a wrong way in chain
2017-06-28 10:39:27 -04:00
fa80177e58
r89: added minimal number of minimizer counts
2017-06-27 18:43:15 -04:00
de54c9dac2
r77: fixed an index loading bug (offset not set)
2017-06-26 13:56:25 -04:00
c01f2dd757
fixed a bug
2017-06-07 11:40:13 -04:00
77c76c1e14
fixed an integer overflow
2017-06-06 22:20:45 -04:00
c1c946aa67
fixed memory leaks
...
though some of them are not really leaks when kalloc is in use.
2017-06-06 21:45:55 -04:00
e6adb673f9
fixed a stupid bug
2017-06-06 15:45:42 -04:00
a2958a4836
minor tweaks
2017-06-06 15:31:21 -04:00
d816e48fce
fixed a bug in chaining
2017-06-06 14:33:43 -04:00
6d4348db44
dp chaining mostly works, but fails sometimes
...
which means there are bugs that need to be fixed
2017-06-06 14:19:50 -04:00
1a9fc04cf0
backup
2017-06-06 10:16:33 -04:00
Heng Li