EPSILON

EPSILON prioritizes expression quantitative trait loci (eQTLs) by using network-based similarity measures computed on local interaction networks to identify the most likely causal gene affecting a target gene's expression when genomic and expression data are associated.

Key Features:

• Local Network Construction: Constructs a local network for each eQTL using a k-trials shortest path algorithm to mitigate effects of highly connected genes and unreliable interactions in physical interaction networks.
• Network-Based Similarity Measures: Evaluates three similarity measures — Random Walks, Laplacian Exponential Diffusion Kernel, and Regularized Commute-Time Kernel — to select the true causal gene.
• Performance Evaluation: Benchmarks performance on three yeast eQTL datasets using different association techniques and compares results to random assignment and shortest path prioritization.
• Enhanced Prioritization Performance: Uses local networks to improve accuracy of predicted knockout interactions from a yeast knockout compendium, reporting an average increase in prioritization performance of 8 percentage points (P < 10^(-5)).
• Implementation: Source code implemented in Matlab/C++.

Scientific Applications:

• eQTL Prioritization: Prioritizing candidate causal genes at eQTL loci to identify regulators of gene expression.
• Causal Gene Identification: Selecting the most likely causal gene among multiple candidates linked to genomic variants.
• Knockout Effect Prediction: Predicting and validating gene knockout interactions using a yeast knockout compendium.
• Gene Regulation Studies: Studying gene regulatory mechanisms and interaction networks in genomic analyses.

Methodology:

Constructs per-eQTL local networks via a k-trials shortest path algorithm; computes Random Walks, Laplacian Exponential Diffusion Kernel, and Regularized Commute-Time Kernel similarity measures; evaluates methods on three yeast eQTL datasets with different association techniques and benchmarks against random assignment and shortest path prioritization; validates predictions using a yeast knockout compendium; implemented in Matlab/C++.