Fingerprint-Enhanced Graph Attention Network (FinGAT) Model for Antibiotic Discovery

Area of Science:

• Computational chemistry
• Machine learning
• Drug discovery

Background:

• Artificial Intelligence (AI) is revolutionizing antibiotic discovery.
• Effective molecular featurization is crucial for accurate AI models in this field.
• Existing Graph Neural Network (GNN) models have limitations in capturing comprehensive molecular information.

Purpose of the Study:

• To develop an advanced AI model for enhanced antibiotic discovery.
• To integrate both sequence-based and structure-based molecular features for improved prediction.
• To introduce the fingerprint-enhanced graph attention network (FinGAT) model.

Main Methods:

• Sequence information is converted into a 2D fingerprint vector.
• Structural information is encoded into a vector using a Graph Attention Network (GAT) module.
• These vectors are concatenated and fed into a multilayer perceptron (MLP) for classification.

Main Results:

• The FinGAT model demonstrated superior performance compared to existing GNN models.
• The integration of sequence and structure features significantly improved antibiotic activity classification.
• FinGAT achieved state-of-the-art results in antibiotic discovery tasks.

Conclusions:

• The FinGAT model offers a powerful new approach for AI-driven antibiotic discovery.
• Combining diverse molecular representations enhances the predictive power of machine learning models.
• This work paves the way for more efficient identification of novel antibiotic compounds.