GENIES

GENIES predicts unknown components of gene and enzyme networks by constructing supervised network inference models that integrate heterogeneous genome-wide data through kernel methods and chemical compatibility constraints.

Key Features:

• Data Integration: Accepts gene expression profiles, protein subcellular localization profiles, phylogenetic profiles, and precomputed gene–gene similarity matrices ("kernels") provided in tab-delimited files.
• Training Data Flexibility: Uses known molecular network information from KEGG PATHWAY or user-provided gene network datasets as training data.
• Algorithm and Parameter Customization: Supports selection of supervised network inference algorithms, adjustment of method parameters, and control of integration weights for heterogeneous data.
• Predictive Output: Produces ranked lists of predicted gene pairs, maps predictions onto KEGG PATHWAY diagrams, and identifies candidate genes for missing enzymes in organism-specific metabolic pathways.
• Chemical Compatibility Integration: Incorporates chemical-compatibility constraints derived from Enzyme Commission (EC) numbers to refine enzyme–enzyme association predictions.

Scientific Applications:

• Metabolic Network Reconstruction: Reconstructs unknown portions of metabolic networks and identifies genes encoding missing enzymes.
• Enzyme Network Inference: Improves prediction accuracy on Saccharomyces cerevisiae datasets through supervised inference and weighted integration of multiple data types.
• Global Enzyme Network Prediction: Enables comprehensive prediction of global enzyme networks by evaluating candidate proteins for enzymatic roles across an organism.

Methodology:

Supervised graph inference using kernel methods to integrate multiple genomic datasets combined with chemical-compatibility constraints derived from Enzyme Commission (EC) numbers; supports algorithm selection, parameter tuning, and weighting of heterogeneous data.