GRADE and X-GRADE: Unveiling Novel Protein-Ligand Interaction Fingerprints Based on GRAIL Scores

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Nonbonding molecular interactions are crucial for biological processes.
• Accurate modeling of these interactions is vital for computational drug design.
• Existing methods may not fully capture interaction quality.

Purpose of the Study:

• Introduce GRADE, a novel interaction fingerprint (IFP) descriptor.
• Quantify nonbonding interactions using floating-point values derived from GRAIL scores.
• Develop a descriptor encoding both the presence and quality of molecular interactions.

Main Methods:

• Developed GRADE, a novel interaction fingerprint (IFP) descriptor.
• Created two variants: a 35-element basic version and a 177-element extended version.
• Utilized Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction.

Main Results:

• GRADE demonstrated competitive performance in dimensionality reduction for protein-ligand complex visualization.
• Achieved reasonable accuracy in binding affinity prediction with minimal machine learning optimization.
• Enhanced the performance of Morgan Fingerprints in 3D-QSAR modeling for a specific protein target.

Conclusions:

• GRADE is a versatile and effective descriptor for analyzing molecular interactions.
• The IFP shows utility in various computational drug design applications, including visualization, prediction, and QSAR.
• GRADE offers a valuable tool for understanding and leveraging nonbonding interactions in drug discovery.