Chiral Cliffs: Investigating the Influence of Chirality on Binding Affinity

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Structural Biology

Background:

• Chirality is often viewed as binary, but its impact on molecular recognition, particularly protein-ligand binding, is more complex.
• The bioactive conformation of chiral molecules is frequently undefined, complicating the prediction of binding activity differences between enantiomers.

Purpose of the Study:

• To investigate chiral behavior in protein-ligand binding and determine when chirality significantly affects binding activity.
• To develop predictive models for identifying 'chiral cliffs'—cases where enantiomers display distinct binding properties.

Main Methods:

• Defined novel 2D descriptors to characterize chiral centers.
• Employed machine learning algorithms to build predictive models using approximately 3800 chiral pairs from the ChEMBL23 database.
• Trained and tested models to discriminate between chiral cliffs and non-cliffs.

Main Results:

• Achieved up to 75% accuracy in predicting chiral cliffs, demonstrating good model performance.
• Developed practical guidelines for deciding whether to use racemic mixtures or enantiopure compounds in binding assays.
• Provided a dataset, novel descriptors, and Python code for model replication.

Conclusions:

• Machine learning models can effectively predict significant differences in binding activity between enantiomers.
• The developed guidelines aid researchers in optimizing experimental design for chiral compounds.
• This work enhances the understanding of chirality's role in molecular recognition and drug development.