PLA-MoRe: A Protein-Ligand Binding Affinity Prediction Model via Comprehensive Molecular Representations

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Accurate prediction of protein-ligand binding affinity is vital for efficient drug discovery.
• Current computational methods often overlook crucial bioactive properties, focusing solely on structural information.
• Wet laboratory experiments for binding affinity determination are costly and time-intensive.

Purpose of the Study:

• To develop a novel computational model, PLA-MoRe, for predicting protein-ligand binding affinity.
• To enhance compound representation by incorporating both structural and bioactive properties.
• To improve the accuracy and reliability of binding affinity predictions in drug discovery.

Main Methods:

• Developed a Graph Isomorphism Network-based structure feature extractor for molecular graphs.
• Designed an Autoencoder-based bioactive feature extractor integrating multisource information (chemical, target, network, cellular, clinical).
• Constructed a sequence feature extractor for protein sequences and combined outputs for prediction via a fully connected network.

Main Results:

• The PLA-MoRe model demonstrated competitive and reliable performance compared to state-of-the-art methods.
• Ablation studies confirmed the contribution of each model component.
• Attention visualization suggested the model's potential in identifying protein binding sites, aiding mechanistic understanding.

Conclusions:

• PLA-MoRe offers a robust approach to protein-ligand binding affinity prediction by leveraging diverse molecular and protein features.
• The model's ability to integrate structural and bioactive information enhances predictive accuracy.
• PLA-MoRe provides a valuable tool for accelerating drug discovery research.