The trick with problem sets so sparsely filled with noteworthy information is skipping as many processing steps as possible without marring the final results by skipping something important.

I have made some modifications to GFS and am testing those now:

• If MSMSFit goes through all of the y-ions and finds less than 12% match then that sequence is skipped
• seqTagScore doesn’t send a sequence to HMM_score unless it has at least 50% match from MSMSFit.
• Sequences weighing less than 1000Da are skipped. This has been changed from 600. This change eliminates 373 additional spectra from consideration from our set of 3010^* which should produce a modest 13% improvement, so if is deemed that we should revert back to 600 the performance hit won’t be devastating.

UPDATE: The job ran 3475 spectra in 1 hour 23
minutes. With efficient load distribution in our pipeline (assuming 12
machines) we should be able to get about 45,000 SPD (spectra per day).

^*3475 is reduced to 3010 when we eliminate spectra with precursor mass less than 600 Da.