HGIMC

HGIMC predicts novel drug-disease associations for computational drug repositioning by combining heterogeneous graph inference with bounded matrix completion and Gaussian radial basis function similarity refinement.

Key Features:

• Heterogeneous Graph Inference: Applies heterogeneous graph inference for network-based association prediction, noted in the input for high convergence precision and rapid speed.
• Matrix Completion (BMC): Uses a bounded matrix completion (BMC) model to prefill missing entries in the drug-disease association matrix and add positive/formative edges between drug and disease networks.
• Improved Similarity Measures (GRB): Employs Gaussian radial basis function (GRB) to refine drug and disease similarity matrices used in inference.
• Novel Network Construction: Constructs a heterogeneous drug-disease network from updated associations and refined similarity metrics to serve as the basis for inferring scores of unknown association pairs.
• Performance Evaluation: Evaluates prediction performance against five state-of-the-art approaches using 10-fold cross-validation and de novo tests, reporting superior prediction performance and computational efficiency in the input.
• Practical Validation: Validates predicted associations via case studies demonstrating utility for identifying potential drug indications.

Scientific Applications:

• Drug Repositioning: Predicts candidate new indications for approved and novel drugs by identifying putative drug-disease associations.
• Computational Pharmacology and Bioinformatics: Supports network-based analyses to prioritize drug-disease hypotheses for downstream experimental validation.

Methodology:

Bounded matrix completion (BMC) to prefill the drug-disease association matrix; Gaussian radial basis function (GRB) to refine drug and disease similarities; construction of a heterogeneous drug-disease network; heterogeneous graph inference to infer association scores for unknown pairs; evaluation via 10-fold cross-validation and de novo tests against five state-of-the-art methods; case-study validation.