Gene expression network reconstruction by convex feature selection when incorporating genetic perturbations.

Area of Science:

• Systems Biology
• Genomics
• Bioinformatics

Background:

• Gene expression data holds valuable regulatory information for discovering network relationships.
• Network reconstruction algorithms often require perturbations to infer unique regulatory links from gene expression data.

Purpose of the Study:

• To introduce a novel algorithm for gene regulatory network reconstruction.
• To leverage the adaptive lasso for selecting regulatory features and cis-expression Quantitative Trait Loci (cis-eQTL) for network inference.

Main Methods:

• Developed a new network reconstruction algorithm powered by the adaptive lasso.
• Utilized cis-eQTL as independent perturbations for maximum network resolution.
• Compared the algorithm's performance against PC-algorithm, QTLnet, QDG, and NEO algorithms.

Main Results:

• The adaptive lasso algorithm outperformed other methods for smaller networks (10 genes, 10 cis-eQTL).
• It demonstrated competitive performance with the QDG algorithm for larger, complex networks (30 genes, 30 cis-eQTL).
• Identified novel regulatory relationships in Saccharomyces cerevisiae, including links to TYR1, RCY1, BUB2, JLP1, and PRM7.

Conclusions:

• The novel algorithm effectively reconstructs directed gene regulatory networks.
• It integrates feature selection with graphical model theory for enhanced network discovery.
• This approach has the potential to reveal new regulatory relationships from population-level genetic and gene expression data.