Metod för klassificering av komplexa prover med låg-upplösnings separationsteknik

Patentradars sammanfattning

The invention provides a classification method for complex mixtures analyzed by low-resolution separation techniques (e.g., SEC, electrophoresis, spectroscopy) where individual components cannot be resolved. By treating the signal envelope and its successive derivatives as linearly independent measurement dimensions, and selecting retention-time regions with statistically significant inter-subgroup differences via a threshold on -log10(p-value), the method constructs multivariate discriminant models (e.g., OPLS-DA). Applied to SEC of blood plasma for Alzheimer's disease versus healthy classification, the combined envelope plus first-derivative approach with optimized thresholds achieves a mean ROC AUC of ~0.980–0.983, substantially outperforming prior-art two-peak methods (~0.877).

AI-genererad och redaktionellt processad.