DRUIDom

DRUIDom predicts interactions between small-molecule compounds and protein structural domains to identify candidate drug-target interactions for drug discovery and repurposing.

Key Features:

• Domain-Based Mapping: Performs statistical mapping of ligands and compounds to protein structural domains to identify putative physical or functional interactions and to transfer associations to other proteins containing the same domains.
• Molecular Clustering and Propagation: Clusters a large-scale set of small-molecule compounds by molecular similarity and propagates identified domain–compound associations to other compounds within those clusters.
• Data-Driven Training: Uses experimentally verified bioactivity data from public databases to build datasets of approximately 2.9 million active/non-interacting compound–target pairs and computes parameters that yield over 27,000 high-confidence associations involving 250 domains and 8,165 compounds.
• Output Generation: Produces approximately 5 million predicted compound–protein interactions derived from propagated domain associations and compound clustering.

Scientific Applications:

• Drug discovery and repurposing: Identifies candidate compound–protein interactions to support target identification and compound prioritization in drug discovery and repurposing projects.
• Experimental validation on LIM-kinases: Has been experimentally validated on LIM-kinase proteins, identifying compounds that inhibit cancer cell migration via inhibition of LIMK phosphorylation and downstream cofilin, including the derivative LIMKi-2d active against drug-resistant Mahlavu liver cancer cells.

Methodology:

Statistical mapping of ligands/compounds to protein structural domains; clustering of compounds by molecular similarity; propagation of domain–compound associations to cluster members; construction of training datasets from experimentally verified bioactivity data (~2.9 million compound–target pairs); calculation of mapping parameters yielding >27,000 domain–compound associations; generation of ~5 million predicted compound–protein interactions.