SPP-CPI

SPP-CPI predicts compound-protein interactions to support drug discovery by integrating distance-matrix compound representations, Spatial Pyramid Pooling networks (SPP-net) for compound feature extraction, and doc2vec protein embeddings.

Key Features:

• Distance Matrix Representation: Represents compounds as inter-atomic distance matrices that capture molecular structural information.
• Spatial Pyramid Pooling Network (SPP-net): Applies SPP-net to compound distance matrices to extract hierarchical spatial features for interaction prediction.
• doc2vec Protein Embeddings: Uses doc2vec to embed protein sequences and capture sequence-level semantic information for protein feature representation.

Scientific Applications:

• Compound–Protein Interaction Prediction: Predicts compound–protein interactions to aid screening and prioritization in drug discovery.
• Benchmark Evaluation: Evaluated on three benchmark datasets: human, C.elegans, and DUDE, where it achieved competitive performance against state-of-the-art predictors.
• Drug-Drug Interaction (DDI) Analysis: Performs DDI experiments demonstrating the utility of distance-matrix representations as molecular characteristics.

Methodology:

Compounds are represented as distance matrices and processed by SPP-net for feature extraction, proteins are embedded with doc2vec, and these features are combined to evaluate compound-protein interactions across benchmark datasets.