To develop biomarkers for neurodegenerative disease that accommodate the co-occurrence of multiple pathological protein species.

We used graph-based clustering on post-mortem histopathological data from 1389 patients with degeneration in the central nervous system to generate 4 non-overlapping, data-driven disease categories that simultaneously account for amyloid-β plaques, tau neurofibrillary tangles, α-synuclein inclusions, neuritic plaques, TDP-43 inclusions, angiopathy, neuron loss, and gliosis. Next, we use cross-validated multiple logistic regression to predict disease labels from in vivo data.

We identified histopathology-based transdiagnostic disease clusters that naturally separated patients in terms of cognitive phenotypes, cerebrospinal fluid (CSF) protein levels, and genotype. This phenotypic separation was not trivially explained by the representation of individual diseases within each cluster. Using multiple logistic regression, we accurately predicted (AUC > 0.9) membership to both existing disease categories and transdiagnostic clusters based on CSF protein levels and genotype.

Broadly, our approach parses phenotypic and genotypic heterogeneity in neurodegenerative disease, and represents a general framework for identifying otherwise-fuzzy disease subtypes in other areas of medicine, such as epilepsy, vascular disease, and cancer. In clinical neurology, our predictive models may be useful for repurposing drugs by comparing efficacy to probabilistic estimates of disease cluster membership, as well as for targeting future drug trials towards an algorithmically defined family of diseases.