Statistical learning for RT-QuIC kinetics phenotyping in prion disease

Background Prion diseases, such as Creutzfeldt-Jakob disease, are invariably fatal neurodegenerative disorders whose clinical heterogeneity makes early diagnosis challenging. The Real-Time Quaking-Induced Conversion (RT-QuIC) assay has revolutionized antemortem detection by amplifying the pathogenic prion protein (PrP Sc ) with near-perfect specificity. However, routine diagnostics typically reduce the continuous RT-QuIC fluorescence curve to just two metrics: the lag phase and the maximum fluorescence intensity. This two-parameter reduction discards the striking heterogeneity of the full temporal profile, erasing latent biological information regarding pathological burden and strain properties. Exploiting this hidden data requires new analytical methods that keep reaction kinetics biologically legible. Methods To address this, we propose an end-to-end statistical framework that captures the complete fluorescence distribution over time. We analyzed cerebrospinal-fluid RT-QuIC trajectories from 301 patients/samples; after exclusion of five patients/samples with consistently shorter recordings, 296 patient/sample-level profiles were included in the tensor modeling and clustering analyses. We then applied a three-way (Tucker3) tensor clustering approach to the unreduced time courses across all technical replicates. Results The proposed framework identifies distinct empirical RT-QuIC fluorescence profiles and provides a compact description of curve morphology. By modelling the complete trajectory, it captures differences in rise steepness, timing, baseline level, and late-phase attenuation that are not represented by conventional lag-phase and maximum-intensity summaries. Conclusions The identified profiles are cohort-specific empirical patterns of the RT-QuIC fluorescence reaction. The late-phase attenuation parameter provides an additional descriptor not included in standard sigmoidal models; however, its biological, clinical, and diagnostic interpretation requires independent validation. Future studies should assess associations between curve-derived parameters or profiles and clinical phenotype, PRNP genotype, molecular subtype, biomarkers, and survival.