EMDomics

EMDomics applies the Earth Mover's Distance to compare gene expression distributions and identify differentially expressed genes in datasets with high within-class heterogeneity.

Key Features:

• Earth Mover's Distance (EMD): Quantifies overall differences in the shape of gene expression distributions between sample classes using the Earth Mover's Distance.
• Permutation-Based q-Values: Derives q-values for each gene via permutations to account for multiple testing and variability within classes.
• Handling Heterogeneity: Captures differences by transforming one distribution into another, addressing high within-class heterogeneity that summary-statistic methods may miss.
• Distribution-Shape Comparison: Provides a more nuanced comparison than methods based solely on summary statistics by evaluating full distributional differences.
• Performance in Simulations: Has demonstrated high sensitivity and specificity for identifying differentially expressed genes in simulated heterogeneous datasets.

Scientific Applications:

• Ovarian cancer drug resistance: Identification of genes associated with drug resistance in ovarian cancer.
• Heterogeneous biomedical datasets: Detection of differential expression in simulated and real-world datasets with heterogeneous sample classes.
• Molecular feature discovery: Investigation of molecular features linked to specific biological or clinical conditions under high intra-group variability.

Methodology:

Computes the Earth Mover's Distance between gene expression distributions, uses permutations to derive q-values for each gene, and assesses distributional transformations between classes.