DEGraph

DEGraph performs hypothesis testing to detect differential expression of predefined gene networks between two experimental conditions by leveraging gene network topology to increase statistical power.

Key Features:

• Topology-aware hypothesis testing: Uses gene network topology to perform more powerful tests for detecting differential expression across gene sets.
• Integration with KEGG pathways: Implements methods to test KEGG pathways for differential expression in gene expression datasets.
• Multivariate two-sample tests of means: Implements multivariate two-sample tests of means to assess location shifts between populations.
• Enhanced power under graph-smooth shifts: Achieves increased statistical power under the assumption that location shifts are smooth on the graph.
• Subgraph identification and multiple testing: Addresses computational challenges and multiple hypothesis testing when identifying nonhomogeneous subgraphs within larger graphs.

Scientific Applications:

• Graph-structured differential expression analysis: Detects shifts in gene expression that align with known graph structures representing biological processes, molecular functions, regulation, or metabolism.
• Cancer gene expression studies: Applied to cancer datasets including breast and bladder cancer using KEGG and NCI pathways.
• Synthetic data evaluation: Illustrated and validated using synthetic datasets to assess performance and robustness.

Methodology:

Performs multivariate two-sample tests of means where the expected location shift correlates with a known graph structure, assumes smooth distribution shifts on the graph, and addresses computational and multiple hypothesis testing challenges for subgraph identification.