Possible false negatives in query about cobs HOT 8 CLOSED

Did you add the --canonicalize flag for cobs compact-construct?

I saw mantis mirrors lexicographically larger k-mers. COBS doesnt by default atm.

from cobs.

thanks for the long bug report, I did find two important errors:
the FastQ reader didnt work with your lines > 64K length. commit 2e79c30
And I fixed the DNA canonicalization. in commit b1720a1

from cobs.

Hope the new version works.

from cobs.

Thanks for your help! I'll let you know how things go

from cobs.

Since commit b1720a1 was only applied to the classic index, can I assume that my compact indexes are correct?

from cobs.

Yes, compacts are built out of classic indexes.

from cobs.

I reran my script and it seems like the results are much closer, but still a bit off. Now, when I query, I get the following

Reading complete index
Read 384.157 MiB / 384.157 MiB - 100%
Index loaded into RAM.
*gb|HQ845196|+|0-861|ARO:3001109|SHV-52 2
ERR1218773      546
ERR1217061      7
TIMER info=search hashes=0.000107381 io=0.000480277 and rows=2.0661e-05 sort results=1.128e-06 total=0.000609447

The greater number of matches for ERR1217061 (7 instead of 4) can probably be explained by false positives, but I'm still not sure why there are only 546 matches to ERR1218773 instead of 642.

from cobs.

Ok, that does indeed fix the problem, I had forgotten to re-enable it during my testing. Thanks for your help!

I'll close this issue then!

from cobs.