A feature selection strategy for gene expression time series experiments with hidden Markov models.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology

Background:

• Transcriptomic time series data often have sparse time points, limiting information extraction.
• Existing methods struggle to effectively analyze such data, necessitating new approaches.

Purpose of the Study:

• To develop a feature selection algorithm for gene expression time course data using a hidden Markov model.
• To address the challenge of sparse sampling in transcriptomic experiments.

Main Methods:

• A two-state hidden Markov model (HMM) was employed for feature selection in gene expression time course data.
• The model assumes no change over time for control samples and at least one change for case samples in case-control studies.
• Features are ranked based on the number and magnitude of changes over time, and replicate quality.

Main Results:

• The proposed method reduced the feature space by up to 90%, retaining relevant genes.
• The strategy proved robust and stable, effectively handling studies with limited sample sizes (e.g., two biological replicates and three time points).
• Findings were consistent with previously reported results from three publicly available datasets.

Conclusions:

• The hidden Markov model-based feature selection is a powerful tool for analyzing sparse transcriptomic time course data.
• This approach overcomes limitations of small sample sizes, enabling more informative gene expression analysis.
• The method enhances the extraction of biologically relevant insights from time-series gene expression experiments.