mapping performance improvement

[alexey0308]

During mapping, stitchWindowAligns uses recursive calls

1. one using the current alignment iA
2. one not using.

The call 1) is performed only if stitchAlignToTranscript does not return a bad score.
Since stitchAlignToTranscript modifies the Transcript object, an additional copy is created, which is expansive.

This pull request attempts to address this performance overhead:
additional cheaper check is performed, which allows to avoid copy-creation of a new Transcript object if it will not be used in the recursion.

The changes comprise of

• adding const modifiers
• extracting a function for transcript finalization (no code modified)
• new function for checking if this alignment can be skipped, used here
• formatting, e.g. long lines of code are split.

Progress output with and without the PR code on a test subset:

           Time    Speed        Read     Read   Mapped   Mapped   Mapped   Mapped Unmapped Unmapped Unmapped Unmapped
                    M/hr      number   length   unique   length   MMrate    multi   multi+       MM    short    other
Nov 02 13:11:28    117.1     2113669      101    94.3%    100.4     0.6%     2.5%     0.1%     0.0%     3.0%     0.0%
ALL DONE!
           Time    Speed        Read     Read   Mapped   Mapped   Mapped   Mapped Unmapped Unmapped Unmapped Unmapped
                    M/hr      number   length   unique   length   MMrate    multi   multi+       MM    short    other
Nov 02 13:13:09     49.5     1168327      101    94.3%    100.4     0.6%     2.6%     0.1%     0.0%     3.0%     0.0%
ALL DONE!

[alexdobin]

Hi Alexey,

thanks a lot for this very interesting PR (and the previous as well).
Sorry for the belayed reply, I am extremely swamped this week and the next one with the Genome Informatics conference. The 2-fold speed-up sounds really big!
Let me do a few checks, and then I will pull it in, next week or the one after.

Cheers
Alex

[rherai]

Does anybody knows whether using a gtf annotation + Star + human genome I could improve alignment speed? If yes, how many times I can speed up the alignment?

[alexdobin]

Hi @rherai

using annotations increases accuracy but not the speed.

Cheers
Alex