Protein Association Neighborhoods Across 37 Taxonomic Groups

Undirected protein association neighborhoods extracted from protein-protein association networks of 1,581 species. It is designed for a 37-way multi-class taxonomic group prediction task as part of the Open Graph Benchmark.

Use Cases

• Train a graph neural network for taxonomic group prediction using the 37-way classification labels.
• Analyze protein association neighborhood structures across the 1,581 represented species.
• Benchmark graph machine learning models on the multi-class classification task defined by the Open Graph Benchmark.

Strengths

• Derived from protein-protein association networks of 1,581 species, providing broad biological coverage.
• Targets a specific 37-way multi-class classification task for standardized benchmarking.
• Part of the curated Open Graph Benchmark, ensuring a defined evaluation framework.

Limitations

• Specific details on graph size, node/edge counts, and feature dimensions are not provided in the input.
• The dataset's creation and preprocessing methodology is only described at an external source.
• The input lacks information on potential class imbalance across the 37 taxonomic groups.

Provenance

Source

Open Graph Benchmark (OGB) team, associated with Stanford.

Collection Method

Extracted from protein-protein association networks.

Freshness

Last updated on the platform was 2023-02-07.