Explainability Methods for Drug-Target Interaction Predictions with GNNs

Mrinal Mahindran's dataset benchmarks four explainable AI attribution methods on graph neural network models for drug-target interactions. The data, last updated in April 2026, includes validation of biological relevance by mapping attributed ligand atoms to 3D protein-ligand structures. It focuses on kinase and G-protein-coupled receptor targets.

Use Cases

• Benchmarking explainable AI methods for graph neural networks based on the described attribution techniques.
• Validating the biological relevance of model predictions by mapping attributed atoms to 3D protein-ligand structures.
• Identifying chemically meaningful ligand features in kinase-inhibitor and GPCR complexes based on the described consensus attributions.

Strengths

• Validates biological relevance with 3D structural mapping, reporting up to 76% of consensus-attributed atoms within 2 Å of the binding pocket.
• Benchmarks four distinct explainable AI attribution methods on GNN models for two specific target classes (kinases and GPCRs).
• File size is 7.7 MB, indicating a manageable download and analysis footprint.

Limitations

• Row count and column-level documentation are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality and structure require manual inspection after download.

Provenance

Source

figshare

Collection Method

Likely contains results from benchmarking explainable AI methods on trained GNN models, as described.

Freshness

Last updated 2026-04-15 19:07:37; freshness should be verified.