Science
Two new papers for high accuracy protein ID, HMM_Score and PMM
01/16/08 17:03
We've been working hard lately on a couple of new
methods for improving the accuracy of protein
identification, and both of these efforts have
recently been published in some good journals.
HMM_Score is based on machine learning principles using a hidden Markov model to significantly improve the accuracy of peptide identification for tandem mass spectrometry (MS/MS) data. The Bioinformatics article is available via Open Access. (DOI 10.1093/bioinformatics/btn011). The associated software can be downloaded from our Downloads page, and updates seen at the HMM_Score page.
The Peptide Markov Model (which, it turns out, doesn't really need the "Markov" part to work well) also uses machine learning principles to significantly improve the accuracy of protein identification using peptide mass fingerprinting (PMF). While many have come to see the PMF approach as out of date due to the advent of ever-better MS/MS based methods, the paper shows that when peptide ion peak intensities are correlated with the peptide sequence, PMF can rival the accuracy of MS/MS based identification. The paper is available via "Author Choice" (open access) here, DOI: 10.1021/pr070088g. We are continuing to explore how this may be used to improve results from shotgun searches as well.
HMM_Score is based on machine learning principles using a hidden Markov model to significantly improve the accuracy of peptide identification for tandem mass spectrometry (MS/MS) data. The Bioinformatics article is available via Open Access. (DOI 10.1093/bioinformatics/btn011). The associated software can be downloaded from our Downloads page, and updates seen at the HMM_Score page.
The Peptide Markov Model (which, it turns out, doesn't really need the "Markov" part to work well) also uses machine learning principles to significantly improve the accuracy of protein identification using peptide mass fingerprinting (PMF). While many have come to see the PMF approach as out of date due to the advent of ever-better MS/MS based methods, the paper shows that when peptide ion peak intensities are correlated with the peptide sequence, PMF can rival the accuracy of MS/MS based identification. The paper is available via "Author Choice" (open access) here, DOI: 10.1021/pr070088g. We are continuing to explore how this may be used to improve results from shotgun searches as well.
|
Followup: no baby boom?
09/19/07 17:43
Since the
premise of the original article was related to NIH
grant funding, and not teaching, I stand by my
original argument: that at or near retirement age,
there is a significant withdrawal of involvement by
these folks in the grant-getting game, even if it is
not officially called "retirement"...
Read
More...
Too many scientists?
09/18/07 13:34
Before making
any drastic changes, it seems worthy to look into the
crystal ball and figure out what else is at play that
might alter the supply/demand equation. While it
would be foolish to try to make accurate predictions,
there are some big trends in demographics that
indicate that we are not necessarily in an
equilibrium situation, and that the present
oversupply of people may change....
Read More...