Targeting neurodegeneration: three machine learning methods for G9a inhibitors discovery using PubChem and...

Area of Science:

• Neuroscience
• Computational Chemistry
• Bioinformatics

Background:

• Growing interest in G9a's role in neurological processes necessitates efficient research tools.
• Existing methods for identifying G9a inhibitors can be time-consuming and costly.
• Machine learning offers a promising avenue for accelerating the discovery and development of G9a-related therapeutics.

Purpose of the Study:

• To develop three distinct machine learning models for predicting G9a inhibitor properties.
• To provide researchers with time-efficient and cost-effective computational tools for G9a research.
• To leverage PubChem data and scikit-learn algorithms for enhanced predictive capabilities.

Main Methods:

• Development of three machine learning models using Python and the scikit-learn library.
• Model 1: Gradient Boosting Regressor for predicting G9a inhibitor efficacy.
• Model 2: Extreme Gradient Boosting Classifier for predicting G9a inhibition based on PubChem identifiers (CID_SID).
• Model 3: Random Forest Classifier for predicting G9a inhibition using IUPAC names.

Main Results:

• The Gradient Boosting Regressor achieved a mean relative error of 17.81% for efficacy prediction.
• The Extreme Gradient Boosting Classifier demonstrated 79.7% accuracy, 83.2% precision, and 78.4% ROC for predicting G9a inhibition from PubChem IDs, with a mean accuracy of 82.75% across seven studies.
• The Random Forest Classifier achieved 68.2% accuracy using IUPAC names, identifying key molecular fragments like 'iodo' and 'phenylcarbamate' associated with G9a activity.

Conclusions:

• Machine learning models provide effective and efficient tools for G9a inhibitor research in neuroscience.
• The developed models demonstrate the potential of computational approaches in accelerating drug discovery.
• The study highlights the utility of PubChem data and various ML algorithms for predicting compound efficacy and activity.