Mean Squared Error of predicted MSMS to observed MSMS

by Brian | 7th July 2009

Present thought: As HMMScore is expensive and findLongestCommonSubstring is both expensive and not ideal for eliminating candidates for HMMScore, I am curious how Mean Squared Error (MSE) of a sequence to the observed MSMS data correlates to HMMScore. If there is strong correlation then it is possible that it is a viable (both robust and low computation cost) selector of sequence candidates.

The cost of MSE is O(M + N) where M and N are the cardinality of the MSMS and sequence respectively.

2 Responses to “Mean Squared Error of predicted MSMS to observed MSMS”

  1. Jainab

    Jul 7th, 2009 :

    It’s good idea to try this. Somehow I think we can improve little bit in HMMScore, and I will try that sometimes this week (both in getBinIntensities() and getIonTypes(). Dynamic programming needs framing the problem, we will try that later, but right now computers are here and we should start submitting jobs by some quick fix. So Brian, you try this and I will see HMMScore and by next week we will start submitting some jobs. While we run our jobs, we will think about redesigning our problem (dynaminc programming in HMMScore, Hastable in genome-based peptide matching, etc).

  2. Jainab

    Jul 9th, 2009 :

    I tried in getIonTypes(). But somehow I thought differently and could not improve that way. Since we are considering the adjacent AA, I could not that way. Tomorrow I will try in getBinIntensity.

Leave a Reply

Name (Required)

Email (Required - will not be published)

Website

Message (Required)