Filter feature selection based Boolean Modelling for Genetic Network Inference

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Gene Regulatory Network (GRN) reconstruction is crucial for understanding biological interactions.
• Existing methods struggle with high-dimensional, low-sample gene expression data and computational efficiency.

Purpose of the Study:

• To introduce a novel combined filter feature selection approach for efficient and accurate GRN inference.
• To demonstrate the efficacy of this approach using a Boolean framework and discretized microarray data.

Main Methods:

• Applied ReliefF for initial gene filtering, followed by a min-redundancy max-relevance criterion for further selection.
• Utilized resampling and a Pearson correlation coefficient-based Boolean modeling approach for rule identification.
• Evaluated the method on small- and medium-scale real gene networks.

Main Results:

• The proposed approach outperformed Linear Discriminant Analysis and individual feature selection methods.
• Achieved improved Structural Accuracy with more true positives compared to state-of-the-art methods.
• Demonstrated superior Dynamic Accuracy and computational efficiency.

Conclusions:

• The novel combined feature selection method offers an efficient and accurate solution for GRN reconstruction from gene expression data.
• This approach effectively handles complex biological dynamics and improves upon existing methodologies.