DINIES

DINIES predicts drug–target interaction networks by supervised network inference that integrates chemical structures, drug side effects, amino acid sequences, and protein domains to prioritize potential drug–protein interactions.

Key Features:

• Heterogeneous data integration: Integrates chemical structures, drug side effects, amino acid sequences, and protein domains for network inference.
• Input formats and kernels: Accepts user-provided profiles or precalculated similarity matrices (kernels) supplied as tab-delimited files.
• Supervised algorithms and parameterization: Implements a range of supervised network inference and machine-learning algorithms with user-adjustable parameters and weights for combining data types.
• KEGG integration: Incorporates known interaction information from KEGG and the KEGG DRUG database or user-supplied interaction datasets to enrich predictive models.
• Functional contextualization: Links predicted interactions to KEGG biological pathways, functional hierarchies, and human disease associations for interpretation.
• Prediction focus: Performs supervised prediction of unknown drug–target interactions using integrated similarity data.

Scientific Applications:

• Interaction discovery: Prediction of unknown drug–target interactions for hypothesis generation.
• Drug discovery and development: Prioritization of candidate targets and compounds for experimental validation.
• Off-target inference: Identification of potential off-target interactions by integrating side-effect and similarity profiles.
• Pathway and disease mapping: Mapping predicted interactions to KEGG pathways and human disease associations for functional interpretation at an omics scale.

Methodology:

Applies supervised machine-learning network inference to similarity matrices/kernels (tab-delimited) derived from chemical, phenotypic, sequence, and domain data, with user-specifiable algorithm selection, parameter weights, and optional incorporation of KEGG/KEGG DRUG or user-supplied interaction datasets.