AGRN: accurate gene regulatory network inference using ensemble machine learning methods.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) govern biological processes.
• GRN dysregulation is linked to complex diseases like cancer and diabetes.
• Accurate GRN inference is crucial for gene function elucidation and disease gene prioritization.

Purpose of the Study:

• To develop a novel method for accurate gene regulatory network inference.
• To leverage an ensemble of machine learning algorithms for improved GRN prediction.
• To utilize Shapley Additive Explanations for robust gene importance scoring.

Main Methods:

• Developed AGRN, an ensemble machine learning approach for GRN inference.
• Employed random forest, extra tree, and support vector regressors.
• Integrated Shapley Additive Explanations to calculate gene importance scores.

Main Results:

• AGRN demonstrated superior performance on benchmark datasets from DREAM4 and DREAM5 challenges.
• Shapley value-based importance scores outperformed traditional methods across datasets.
• The ensemble approach with Shapley values enhanced GRN inference accuracy.

Conclusions:

• AGRN provides a more accurate method for gene regulatory network inference.
• The findings suggest AGRN can significantly advance mechanistic understanding of biological processes.
• Improved GRN inference facilitates faster identification of genes implicated in diseases.