Related Experiment Videos
Bayesian classifiers for detecting HGT using fixed and variable order markov models of genomic signatures
Daniel Dalevi1, Devdatt Dubhashi, Malte Hermansson
1Department of Computing Science, Chalmers University, SE 412 96 Göteborg, Sweden. dalevi@cs.chalmers.se
Bioinformatics (Oxford, England)
|January 13, 2006
Summary
Genomic signature analysis using Markov models improves bacterial classification accuracy. This method enhances prediction flexibility and identifies foreign genes, aiding in understanding horizontal gene transfer.
Area of Science:
- Bioinformatics
- Genomics
- Computational Biology
Background:
- Genomic signature analyses offer species-specific insights without sequence alignment.
- A Bayesian classifier successfully identified foreign genes in Neisseria meningitis.
- This study advances classification using Markov models (Mk, VLMk).
Purpose of the Study:
- To enhance genomic signature classification accuracy and flexibility.
- To investigate the performance of variable length Markov models (VLMk).
- To develop methods for hypothesis testing in gene transfer.
Main Methods:
- Implementation of fixed and variable length Markov models (Mk, VLMk).
- Development of an algorithm to parameterize VLMk.
- Evaluation of classifier integrity using false-negative rates and training data size estimation.
Main Results:
- Markov models significantly increase prediction flexibility and accuracy compared to naive models.
- Estimates for minimal training data sizes were determined.
- A method to reject false hypotheses of horizontal gene transfer was proposed.
Conclusions:
- Markov models represent a significant advancement in genomic signature classification.
- The developed methods improve the identification of foreign genes and horizontal gene transfer events.