Spatial chemical distance based on atomic property fields | JoVE Visualize

Area of Science:

Computational chemistry
Drug discovery
cheminformatics

Background:

Compound chemical structure similarity often correlates with similar pharmacological profiles and protein targets.
Distinct chemical scaffolds can also exhibit similar pharmacology, posing challenges for traditional similarity searches.
Relying solely on chemical similarity can limit the discovery of novel drug candidates and hinder lead hopping.

Purpose of the Study:

To design a novel compound similarity/distance measure that better reflects structural aspects of pharmacology and molecular interactions.
To improve the identification of pharmacologically similar compounds and facilitate lead hopping in drug discovery.
To overcome limitations of existing 2D and shape-based similarity measures.

Main Methods:

Developed a new similarity measure based on spatial alignment of compounds using atomic property fields as generalized 3D pharmacophoric potentials.
Optimized atomic property contributions using Partial Least Squares (PLS) regression for enhanced discrimination of pharmacologically similar vs. dissimilar compound pairs.
Validated the measure on a diverse dataset of 115 protein-ligand complexes, comparing performance against 2D Tanimoto and Shape Tanimoto methods.

Main Results:

The proposed spatial chemical distance measure demonstrated improved performance in discriminating pharmacologically similar compound pairs compared to 2D Tanimoto and Shape Tanimoto.
Significant improvements in Area Under the Receiver Operating Characteristic Curve (AUC) values were observed in 66% (vs. 2D Tanimoto) and 58% (vs. Shape Tanimoto) of domains.
The new approach showed particularly high improvement for challenging cases with low initial AUC values (<0.8), achieving 86% and 85% improvement over 2D Tanimoto and Shape Tanimoto, respectively.

Conclusions:

The developed spatial chemical distance measure offers a more effective approach for assessing compound similarity based on pharmacological and interaction profiles.
This method enhances virtual ligand screening by providing better discrimination of active compounds and enabling broader exploration of chemical space.
The findings support the utility of 3D pharmacophoric potentials and spatial alignment for advancing drug discovery and lead optimization efforts.