DeepCPI

DeepCPI predicts compound-protein interactions using deep learning to enable large-scale in silico drug screening and drug repositioning.

Key Features:

• Feature Embedding and Representation Learning: Employs feature embedding techniques to automatically extract implicit yet expressive low-dimensional representations of compounds and proteins from large volumes of unlabeled data.
• Scalability and Generalizability: Architected for generality and scalability across diverse compound and protein types and large datasets.
• Superior Predictive Performance: Demonstrated high predictive accuracy when evaluated against large-scale databases such as ChEMBL, BindingDB, and DrugBank.
• Experimental Validation: Predictions were experimentally validated for small-molecule interactions with glucagon-like peptide-1 receptor, glucagon receptor, and vasoactive intestinal peptide receptor.

Scientific Applications:

• Large-scale in silico screening: Enables high-throughput prediction of compound–protein interactions to prioritize compounds for experimental testing.
• Drug repositioning and target identification: Facilitates identification of novel therapeutic targets and drug repurposing opportunities from large datasets.

Methodology:

Uses deep learning methodologies with feature embedding and representation learning on large volumes of unlabeled data and evaluation against ChEMBL, BindingDB, and DrugBank.