GraphAT

GraphAT performs graph-theoretic association tests to evaluate the significance of associations among biological entities (e.g., genes, proteins) across functional genomics datasets such as Gene Ontology-derived predictomes, mRNA expression profiles, and phenotype data.

Key Features:

• Graph-theoretic framework: Models biological entities as nodes and interactions or associations as edges to represent complex relationships among genes and proteins.
• Edge permutation test: Assesses significance of graph connections by permuting edges while preserving node degrees to test whether observed associations exceed random expectation.
• Node label permutation test: Evaluates significance by randomly shuffling node labels to estimate the likelihood that observed associations arise by chance.
• Integration of diverse data sources: Combines Gene Ontology-derived predictomes, mRNA expression profiles, and phenotype data to test associations across heterogeneous functional genomics datasets.
• Application to Saccharomyces cerevisiae: Demonstrated on S. cerevisiae datasets to identify associations between gene ontology predictions and experimental mRNA expression and phenotypic data.
• Resolution of dataset discrepancies: Enables investigation of discrepancies between data types, such as differences between gene expression profiles and knockout phenotypes, using alternative datasets for validation.

Scientific Applications:

• Systems biology: Testing network-level hypotheses about interactions and functional relationships among genes and proteins within cellular systems.
• Functional genomics: Evaluating associations between predictomes, expression profiles, and phenotypic measurements to infer gene functions.
• Comparative validation: Comparing and validating signals across heterogeneous datasets, including resolving conflicts between expression and knockout phenotype data.
• Yeast genetics studies: Identifying and statistically validating associations in Saccharomyces cerevisiae functional genomics experiments.

Methodology:

Construct a graph of biological entities (nodes) and interactions (edges), then apply permutation-based statistical tests including edge-permutation preserving node degrees and node-label permutation to evaluate association significance.