Data-Driven and Knowledge-Based Algorithms for Gene Network Reconstruction on High-Dimensional Data

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene network reconstruction traditionally relies on data-driven models.
• Integrating prior biological knowledge offers a promising avenue for improved accuracy.
• Small sample sizes and high dimensionality pose significant challenges in gene network analysis.

Purpose of the Study:

• To develop novel algorithms for gene network reconstruction.
• To incorporate prior biological knowledge into network inference.
• To address challenges of small sample sizes and high dimensionality.

Main Methods:

• Utilized empirical Bayesian approach for covariance matrix estimation via shrinkage.
• Employed penalized normal likelihood method for Gaussian graphical model selection.
• Developed algorithms for gene network reconstruction with and without prior knowledge.

Main Results:

• Proposed algorithms demonstrate superior performance compared to state-of-the-art methods.
• Achieved improved Precision-Recall (PR) and Receiver Operating Characteristic (ROC) curves.
• Successfully applied the method to human gastric atrophy RNA-seq data.

Conclusions:

• The developed algorithms effectively reconstruct gene networks, especially when prior knowledge is incorporated.
• The empirical Bayesian and penalized likelihood methods provide a robust framework for network inference.
• This approach offers significant advancements for analyzing complex biological networks from expression data.