Developing predictive models for µ opioid receptor binding using machine learning and deep learning techniques.

Area of Science:

• Computational chemistry
• Pharmacology
• Machine learning

Background:

• Opioids provide pain relief by activating the µ opioid receptor (MOR), but their addictive nature contributes to the ongoing opioid crisis.
• Understanding the relationship between chemical structure and MOR binding is crucial for developing safer analgesics.
• Predictive models can accelerate the discovery of non-addictive or less-addictive opioid-based pain therapeutics.

Purpose of the Study:

• To develop and evaluate machine learning (ML) and deep learning (DL) models for predicting the MOR binding activity of chemical compounds.
• To assess the performance of various ML algorithms in identifying potential MOR ligands.
• To investigate the utility of prediction confidence and applicability domain analyses for model interpretation.

Main Methods:

• Curated a dataset of chemicals with known MOR binding activity from public databases and literature.
• Calculated molecular descriptors for each chemical using Mold2 software.
• Trained and validated multiple ML models, including Random Forest, k-Nearest Neighbors, Support Vector Machine, Multi-layer Perceptron, and Long Short-Term Memory, using 5-fold cross-validation and external validation sets.

Main Results:

• The developed models achieved Matthews correlation coefficients (MCC) ranging from 0.528 to 0.654 in cross-validation and 0.408 in external validation.
• Prediction confidence and applicability domain analyses were identified as important factors for model reliability.
• The models demonstrated potential in identifying chemicals with MOR binding capabilities.

Conclusions:

• The developed ML and DL models show promise for predicting MOR binding activity.
• These predictive tools can aid in the screening and identification of novel MOR binders.
• The findings support the potential development of safer, non-addictive analgesics targeting the MOR pathway.