SMILE

SMILE applies Linear Genetic Programming (LGP) to supervised metabolomics datasets to produce interpretable predictive models and detect metabolite interactions related to disease states.

Key Features:

• Linear Genetic Programming (LGP): Uses Linear Genetic Programming as the core evolutionary learning algorithm for model construction.
• Python implementation: Implements the algorithm within a Python environment.
• scikit-learn-like API: Exposes an API similar to scikit-learn for model training and evaluation.
• Supervised metabolomics analysis: Operates on supervised metabolomics datasets to perform classification tasks.
• Feature interaction discovery: Focuses on identifying feature interactions and influential metabolite combinations through data mining.
• Binary classification focus: Is tailored for binary classification scenarios common in case-control metabolomics studies.
• Interpretability via evolution: Employs an evolutionary approach that evolves models optimized for interpretability alongside predictive performance.
• Interpretable predictive models: Outputs models that elucidate metabolite contributions and their interactions relevant to biological states.

Scientific Applications:

• Alzheimer's disease (AD) research: Analyzes metabolomics data to identify key metabolites implicated in AD and produce interpretable predictive models.
• Biomarker discovery and classification in metabolomics: Supports discovery of interacting metabolite signatures and predictive models in case-control studies.

Methodology:

Linear Genetic Programming (LGP) implemented in Python with a scikit-learn-like API; an evolutionary algorithm evolves symbolic models for supervised learning (primarily binary classification) and identifies interacting features for interpretability.