StochHMM: a flexible hidden Markov model tool and C++ library

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning

Background:

• Hidden Markov models (HMMs) are powerful probabilistic tools for classification in computational biology.
• Traditional HMM implementations often lack flexibility for advanced research needs.

Purpose of the Study:

• To introduce StochHMM, a C++ library and command-line program for flexible Hidden Markov Model implementation.
• To provide researchers with enhanced capabilities for HMMs in biological data analysis.

Main Methods:

• StochHMM is implemented in C++.
• It supports traditional HMMs from simple text files.
• The framework allows for higher-order emissions and integration of user-defined functions and data sources.

Main Results:

• StochHMM enables user-defined alphabets and handling of ambiguous characters.
• Researchers can implement user-defined weighting of state paths.
• Transition probabilities can be tied to sequences, offering greater model customization.

Conclusions:

• StochHMM provides a flexible and extensible platform for applying Hidden Markov Models in computational biology.
• The tool empowers researchers to tailor HMMs for specific and complex classification problems in biological data.