EnsInfer: a simple ensemble approach to network inference outperforms any single method.

Area of Science:

• Computational Biology
• Bioinformatics
• Statistical Inference

Background:

• Causal inference methods are crucial for analyzing complex biological networks, particularly RNA sequencing data.
• Existing base causal inference methods exhibit variable performance across different datasets.
• The development of robust ensemble methods is needed to improve the reliability of causal inference.

Purpose of the Study:

• To evaluate the performance of various base causal inference methods and ensemble approaches.
• To introduce and detail the EnsInfer ensemble model for integrating multiple causal inference techniques.
• To demonstrate the effectiveness of EnsInfer using RNA-seq data.

Main Methods:

• Evaluation of diverse base causal inference algorithms.
• Development and application of a non-homogeneous ensemble classifier (Naive Bayes).
• Integration of methods passing a statistical normality test on training data.

Main Results:

• Base network inference methods show dataset-dependent performance.
• The Naive Bayes ensemble classifier achieved results comparable to or better than the best single base method.
• EnsInfer successfully integrates various RNA-seq data types and inference methods.

Conclusions:

• Ensemble methods, particularly EnsInfer, offer superior and more robust causal inference than single methods.
• EnsInfer provides a flexible framework for integrating diverse inference techniques and data types.
• The developed EnsInfer model and associated data will be publicly available.