NESS

NESS integrates heterogeneous functional genomics data, ontologies, annotations, and curated experimental results into multi-species graph networks and applies diffusion metrics, including random walk with restart (RWR), to quantify relationships among genes, diseases, and phenotypes.

Key Features:

• Multi-Species Data Integration: Aggregates data from multiple species to address sparsely populated and biased gene annotation datasets.
• Graph-Based Harmonization: Harmonizes ontologies, annotations, and curated experimental results into a cohesive graph-based model for cross-species analysis.
• Diffusion Metrics Utilization: Employs diffusion metrics, specifically random walk with restart (RWR), to estimate relationships among entities within the integrated network.
• Resistance to Data Sparsity and Noise: Designed to be resistant to spurious or sparse datasets, improving recapitulation of ground truth biological pathways compared to other similarity metrics alone.
• Integration with GeneWeaver: Embedded in the GeneWeaver environment to leverage curated multi-species networks and enable assertions about gene-gene, gene-disease, and phenotype-disease relatedness.

Scientific Applications:

• Disease Concept Classification: Refines classification of biological concepts, particularly disease, by evaluating large-scale genomic datasets.
• Bias Mitigation and Dataset Augmentation: Mitigates experimental biases and augments sparse datasets through cross-species analysis.
• Animal Model Integration: Integrates animal model studies to enhance understanding of gene function and biological processes in health and disease contexts.
• Substance Use Disorder Research: Supports identification of genes and biological pathways with shared associations across co-occurring disorders, including substance use disorders.

Methodology:

Harmonization of heterogeneous functional genomics data into a unified graph-based model and application of the random walk with restart (RWR) diffusion algorithm to score relationships among genes, processes, and diseases through shared genomic associations, reducing bias from sparse or noisy datasets.