Machine-OlF-Action

Machine-Olf-Action applies machine learning to classify chemical molecules and prioritize biologically relevant agonists for chemosensory G-Protein Coupled Receptors (GPCRs).

Key Features:

• SMILES input and annotations: Accepts SMILES representations and activation-status annotations to construct classification datasets.
• Chemical database integration: Integrates chemical databases totaling approximately 103 million chemical moieties for large-scale screening.
• Customized screening: Performs customized screening on supplied datasets to evaluate chemosensory interactions and molecular properties.
• Machine learning classification: Trains classification models using machine learning algorithms on annotated chemical data.
• Model interpretability with LIME: Implements Local Interpretable Model-agnostic Explanations (LIME) to produce local neighborhood-based molecular embeddings and interpret classifier predictions.

Scientific Applications:

• Agonist discovery for olfactory receptors: Identification of previously unreported agonists for human OR1A1 and mouse MOR174-9.
• Prediction of biologically active compounds: Leverages chemical feature patterns from known agonists and non-agonists to predict novel bioactive molecules affecting GPCR-associated chemosensory mechanisms.
• Large-scale compound screening: Enables screening and analysis of extensive chemical compound collections using integrated databases and classification models.

Methodology:

Machine-Olf-Action trains machine learning classification models from SMILES and activation-status annotations, integrates chemical databases for screening, and applies LIME to generate local neighborhood-based molecular embeddings and interpret predictions.