A Structure-Based Platform for Predicting Chemical-Induced Parkinson's Disease.

Area of Science:

• Computational toxicology
• Pharmacology
• Neuroscience

Background:

• Current methods for assessing agrochemical neurotoxicity are insufficient, evidenced by links between pesticide exposure and Parkinson's disease (PD).
• Mechanism-based in silico screening offers a predictive approach for molecular initiating events, crucial precursors to adverse outcomes.
• Limited protein-binding data for pesticides necessitates extrapolation methods for broad chemical screening.

Purpose of the Study:

• To develop and validate a mechanism-based in silico screening approach for predicting neurotoxic potential.
• To utilize group I metabotropic glutamate receptors (mGluRs) as a case study due to their involvement in chemical-induced PD.
• To create a user-friendly platform for implementing the developed screening models.

Main Methods:

• Docking of known active compounds into the allosteric binding site of mGluRs.
• Computation of interaction fingerprints (IFPs) and training of classification models.
• Evaluation of model enrichment (ROC AUC), feature importance, and applicability domain analysis.

Main Results:

• IFP-based mGluR models showed good predictive enrichment (ROC AUC 0.78 and 0.66).
• Key interactions identified included hydrogen bonds with Asn760 (mGluR1) and aromatic interactions with Trp785 (mGluR5).
• Virtual screening identified 132 potential mGluR binders, including bifenthrin, validating the model's efficacy.

Conclusions:

• Interaction fingerprints combined with machine learning provide a powerful tool for mechanism-based in silico toxicology.
• The developed screening technique can predict potential neurotoxicants and aids in the development of safer agrochemicals.
• This approach contributes to a paradigm shift towards predictive toxicology, reducing reliance on traditional testing methods.