Symmetric time warping, Boltzmann pair probabilities and functional genomics

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Classical time warping, used in speech recognition, lacks symmetry, limiting its application in biological time series analysis.
• Existing methods may not accurately reflect the relationship between time series of different lengths or orientations.

Purpose of the Study:

• To introduce a novel symmetric time warping algorithm for improved time series alignment.
• To provide a formal proof of symmetry for the new algorithm and a variant of Aach and Church.
• To develop dynamic programming algorithms for computing Boltzmann partition functions in symmetric time warping.

Main Methods:

• Designed a new symmetric time warping algorithm.
• Formally proved the symmetry of the proposed algorithm and a variant.
• Developed quadratic time dynamic programming algorithms for Boltzmann partition functions.

Main Results:

• Established a formally proven symmetric time warping algorithm.
• Enabled computation of Boltzmann probabilities for time series point alignments.
• Demonstrated potential applications in analyzing gene expression data.

Conclusions:

• The new symmetric time warping algorithm offers enhanced accuracy for time series alignment.
• This method provides a framework for inferring biological significance from aligned time points in gene expression data.
• The publicly available web server facilitates further research and application in bioinformatics.