MolTrans

MolTrans predicts drug–target interactions by combining knowledge-inspired sub-structural pattern mining with an augmented transformer encoder to extract semantic relations among molecular sub-structures for more accurate and interpretable DTI prediction.

Key Features:

• Sub-Structural Pattern Mining: A knowledge-inspired algorithm that identifies and leverages molecular sub-structures to improve prediction accuracy and interpretability.
• Interaction Modeling Module: A module that models nuanced relationships between drug compounds and protein targets to enhance DTI prediction precision.
• Augmented Transformer Encoder: A transformer-based encoder that processes massive unlabeled biomedical data to capture semantic relations among sub-structures and improve learning from limited labeled datasets.

Scientific Applications:

• Drug–Target Interaction Prediction: Predicts interactions between small molecules and protein targets to support identification of potential therapeutic compounds.

Methodology:

The methodology comprises a sub-structural pattern mining algorithm and an augmented transformer encoder that processes massive amounts of unlabeled biomedical data to capture semantic relationships among molecular sub-structures.