The good news is that preliminary performance of my MSE algorithms is FAST… like, doesn’t even add a second to overall performance. If I can get some good correlation between HMMScore and MSE (especially low scoring HMMScore) then it could be a really good filter. Right now there is not good correlation, but I feel it is because my algorithm is not taking into account certain edge cases or perhaps I’m not totally understanding what makes a sequence a good candidate for a decent HMMScore.

In that vein, I’ve output visualizations of some acid sequences compared to MSMS spectrum data. The first sequence is the “real” sequence (the one predicted by GFS). The rest are candidates.

MMDVIQEIER

VAFGTYCRMR

MTPANDDDVKR

QPKIESDCQR

ASECREQPGVK

GVECQAQIAER

YGAGQQTNLFY.

GNTLLWQEDW.

VTGHGFGCRNR

GKDCRFHGQR