Entering edit mode
7.0 years ago
Tamador
▴
10
I worked with group in project about how we can use hidden Markov model to detect more member of gene family using viterbi algorithm . We construct HMM from DNA aligned sequences and then we used viterbi algorithm to calculate score for query sequence.
My question is how I can use z-score to decide if the results from viterbi are significant or not ? should I calculated the mean and the Standard deviation from aligned sequences ? if yes, how i do that and if no , what I should do ?
I know this is not your question, but the choice to use the Viterbi algorithm for detecting additional members of a gene family strikes me as odd. The "standard" approach would be to use the forward-backward algorithm to calculate the posterior probability of the query sequence given the HMM.
Sorry for the inconvenience , but could you tell me what is the different between viterbi and forward algorithm ??
Ignoring the issue of whether to use the Viterbi or forward-backward algorithm, I would in any case not use z-scores for assessing significance. The relationship between z-score and statistical significance is based on assuming a normal distribution. I would not expect the scores from the Viterbi algorithm to follow the normal distribution; my best guess would be that they follow an extreme value distribution.
Our project was based on the use of viterbi algorithm and z score , that's not my choice
What software are you using to construct the HMM? HMMER will take care of most of the issues for you: it will use forward-backward and EVD to estimate the significance.
I did not use any software , we wrote a c++ code to build HMM