Harmonizome

Harmonizome integrates knowledge about genes and proteins from diverse multi-omics resources, enabling researchers to explore and integrate biological data through unified gene-attribute associations.

Key Features:

• Multi-omics Data Integration: Processes and abstracts diverse omics datasets into associations between genes and their attributes from many central omics resources.
• Dataset Crossing: Identifies gene modules that are shared across datasets.
• Large Language Model Explanation: Uses a large language model to compose paragraphs speculating about reasons behind significantly high gene set overlap between dataset pairs.
• Knowledge Graph Assertions: Provides datasets downloadable as knowledge graph assertions.
• UMAP Visualization: Visualizes datasets with Uniform Manifold Approximation and Projection plots.
• Ball-and-stick Diagrams: Visualizes gene-attribute associations as ball-and-stick diagrams.
• AI-ready Formats: Provides processed omics datasets in several AI-ready formats.

Scientific Applications:

• Gene-Attribute Association Exploration: Enables researchers to explore associations between genes and attribute types such as cells and tissues, diseases, and pathways.
• Gene Module Identification: Identifies shared gene modules across different datasets through dataset crossing.
• Gene Set Overlap Analysis: Explains significantly high gene set overlap between dataset pairs using automated LLM-generated speculation.

Methodology:

Processes and abstracts diverse omics datasets into associations between genes and their attributes; uses a large language model to compose explanatory paragraphs for high gene set overlap; applies Uniform Manifold Approximation and Projection for visualization.