FRnet-DTI

FRnet-DTI predicts drug-target interactions using FRnet-Encode auto-encoder features and a FRnet-Predict convolutional neural network classifier to support pharmacology and therapeutic drug design.

Key Features:

• Auto-encoder feature extraction (FRnet-Encode): FRnet-Encode is an auto-encoder model that generates a comprehensive 4096-dimensional feature vector for each instance.
• Convolutional neural network classifier (FRnet-Predict): FRnet-Predict is a CNN classifier that uses the 4096 features to predict interaction probabilities between drugs and targets.
• Performance evaluation on benchmark datasets: The method was evaluated on four gold standard datasets and demonstrated improvements over existing methods in area under the Receiver Operating Characteristic curve (auROC) and area under the Precision-Recall curve (auPR).
• Novel interaction identification: The approach identified twenty new potential drug-target pairs with high predicted interaction probabilities.

Scientific Applications:

• Drug-target interaction prediction: Generate probabilistic predictions of interactions between small-molecule drugs and protein targets to inform pharmacology studies.
• Therapeutic design and repurposing: Support therapeutic drug design and the prioritization of candidate compounds for repurposing by identifying high-probability drug-target pairs.

Methodology:

FRnet-Encode uses an auto-encoder to transform input data into a 4096-dimensional feature space, and FRnet-Predict applies a convolutional neural network to those features to compute drug-target interaction probabilities.