FraGAT: a fragment-oriented multi-scale graph attention model for molecular property prediction

Area of Science:

• Computational Chemistry
• cheminformatics
• Drug Discovery

Background:

• Molecular property prediction is crucial in drug discovery.
• Current graph-based models overlook molecular hierarchical structures.
• Functional groups significantly influence molecular properties and binding affinities.

Purpose of the Study:

• To develop a novel model for enhanced molecular property prediction.
• To leverage molecular fragments containing functional groups for improved predictions.
• To introduce a fragment-oriented multi-scale graph attention network (FraGAT).

Main Methods:

• Defined molecule graph fragments incorporating functional groups.
• Developed a fragment-oriented multi-scale graph attention network (FraGAT).
• Evaluated FraGAT on widely used benchmarks for molecular property prediction.

Main Results:

• FraGAT achieved state-of-the-art predictive performance across multiple benchmarks.
• Models using functional group-containing fragments demonstrated superior predictive capabilities.
• Case studies confirmed the model's interpretability and effectiveness.

Conclusions:

• FraGAT effectively predicts molecular properties by utilizing functional group information.
• The fragment-oriented approach enhances model performance and interpretability.
• This method offers a promising direction for advancing molecular property prediction.