Back

Theta, a multidimensional ratio biomarker applied to five amyloid beta peptides for investigations in familial Alzheimer's disease

Llorente Saguer, I.; Arber, C.; Oxtoby, N. P.

2025-08-08 neurology
10.1101/2025.08.06.25333131 medRxiv
Show abstract

Biomarker discovery in complex diseases often requires classifying biological states using multiple, interconnected features, a task where traditional two-feature ratios are limited. We introduce theta, a mathematical model designed to extend the ratio concept for classification across any number of input features. Theta is a multivariate normative statistical model, which quantifies the deviation of a multidimensional feature vector relative to a reference. We apply theta to amyloid-beta (A{beta}) peptide profiles to classify early-onset familial Alzheimers disease mutation carriers versus controls using multiple datasets of in vitro neuronal models. Theta consistently achieved high discrimination (AUC > 0.99, precision-recall AUC > 0.96) across multi-site datasets, considerably outperforming established A{beta} biomarkers (precision-recall AUC 0.41 is the best score from the literature, in all datasets merged). This demonstrates thetas power as a versatile mathematical tool.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above