InceptionDTA: Predicting drug-target binding affinity with biological context features and inception networks.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Accurate drug-target binding affinity prediction is vital for efficient drug discovery.
• Traditional machine learning and existing deep learning models face limitations in feature extraction and scalability.

Purpose of the Study:

• To introduce InceptionDTA, a novel deep learning model for predicting drug-target binding affinity.
• To address limitations of existing models in capturing biological context and multi-scale features.

Main Methods:

• Developed InceptionDTA, utilizing CharVec for enhanced protein sequence encoding with biological context.
• Employed a multi-scale convolutional architecture inspired by the Inception network for feature extraction from protein sequences and drug SMILES.
• Evaluated performance across benchmark datasets using warm-start, refined, and cold-start settings.

Main Results:

• InceptionDTA significantly outperformed sequence-based, transformer-based, and graph-based deep learning methods.
• The CharVec-enhanced version achieved high accuracy in absolute predictions.
• A label-encoding version demonstrated strong performance in ranking and predicting relative binding affinities.

Conclusions:

• InceptionDTA offers a versatile and effective approach for drug-target binding affinity prediction.
• The model shows promise in accelerating drug repurposing and facilitating new drug discovery.
• This work contributes to advancing computational methods for disease treatment.