genra-py

genra-py implements the Generalized Read-Across (GenRA) methodology to estimate physico-chemical, biological, and eco-toxicological properties of chemicals by inferring values from analogous substances for computational toxicology and chemical safety assessment.

Key Features:

• Data-driven approach: Employs similarity-weighted activity assessments to infer chemical properties from analogue compounds.
• Interpretable, automated read-across: Automates read-across while systematically exploring input data selection and neighborhood definitions to enable objective evaluation of predictive performance.
• Nearest-neighbor retrieval: Uses nearest-neighbor algorithms to identify analogous chemicals for information transfer.
• scikit-learn compatibility: Adheres to the scikit-learn estimator design pattern for use within machine-learning workflows.

Scientific Applications:

• Chemical safety analysis and risk assessment: Provides analogue-based predictions to address data gaps for new or untested chemicals.
• Hazard identification and point of departure estimation: Supports human health risk assessment tasks by supplying read-across estimates for hazard characterization and POD derivation.
• Regulatory toxicology and environmental science: Supplies computational evidence for decision-making in regulatory assessments and environmental evaluations.

Methodology:

Nearest-neighbor algorithms identify analogous chemicals and similarity-weighted activity assessments generate read-across estimates, with systematic exploration of input data selection and neighborhood definition; the package follows the scikit-learn estimator design pattern.