fishy.analysis.biomarker

Automated Biomarker Discovery module.

This module maps high-importance m/z values (from XAI) to chemical entities using public databases (LipidMaps).

Classes

class fishy.analysis.biomarker.AutomatedBiomarkerDiscovery(m_z_values: List[float])

Bases: object

Connects XAI results to chemical databases for automated biomarker identification.

__init__(m_z_values: List[float]) → None

Parameters:

m_z_values (List[float]) – List of m/z values corresponding to the input features.

generate_report(biomarker_results: List[Dict[str, Any]]) → str

Formats the discovery results into a readable string.

identify_top_peaks(importance_scores: ndarray, top_n: int = 5) → List[Dict[str, Any]]

Finds the top N m/z values and queries databases.

query_lipidmaps(mz: float, tolerance: float = 0.1) → List[Dict[str, Any]]

Queries LipidMaps REST API for potential matches for a given m/z. Uses M-H ion type for negative mode REIMS.

Functions

fishy.analysis.biomarker.run_biomarker_pipeline(model, data_loader, feature_names, device, top_n=5)

Complete flow: Model -> Grad-CAM -> Top Peaks -> Chemical Identity

s