<>
Researchers from the University of Miami and the UK Biobank successfully tracked pre-symptomatic protein changes in plasma samples to forecast phenoconversion—the transition from carrying a genetic mutation to developing active disease—offering a critical new window for early therapeutic intervention in familial and sporadic ALS cases.
Identifying Pre-Symptomatic Protein Biomarkers
To detect molecular changes before physical weakness manifests, the research team analyzed 516 plasma samples using Olink’s Explore HT proximity extension assay technology, according to the study data. The discovery cohort drew participants from three primary initiatives: the Pre-fALS study (ClinicalTrials.gov Identifier: NCT00317616), the Clinical Research in ALS (CRiALS) Biomarker study (NCT00136500), and the CReATe Consortium Phenotype-Genotype-Biomarker study (NCT02327845). By comparing healthy controls with patients exhibiting clinically manifest ALS, investigators identified a core panel of differentially expressed proteins associated with disease pathology.
The institutional review board at the University of Miami approved all protocols, and every participant provided written informed consent. Plasma was collected, centrifuged at 1,750g for 10 minutes at 4 degrees Celsius, and stored at minus 80 degrees Celsius following strict standard operating procedures. Using generalized additive mixed models (GAMMs), the team mapped longitudinal protein trajectories to determine the earliest timeline of pre-symptomatic divergence from age- and sex-matched healthy controls.
Machine Learning Models for Phenoconversion Timing
The study applied machine learning algorithms, specifically logistic regression and partially penalized Cox proportional hazards models with LASSO regularization, to predict both the occurrence and exact timing of phenoconversion. According to the findings, the multi-protein panel significantly improved estimation of proximity to symptom onset compared to models relying solely on neurofilament light chain (NEFL) levels.
Using absolute protein expression values (NPX) rather than relative abundance enabled the creation of risk scores that function independently of a reference control population.
Independent Replication in the UK Biobank Cohort
To validate the discovery models externally, researchers turned to the UK Biobank, analyzing data from over 52,000 participants with available whole-genome sequencing and Olink Explore 3072 proteomic assays. Replication analyses confirmed the directional consistency of overlapping protein markers between the University of Miami discovery cohort and the large-scale UK population data.
While the discovery cohort utilized longitudinal blood draws, the UK Biobank replication relied on cross-sectional generalized additive models (GAMs) due to the nature of the database. The successful cross-cohort replication demonstrates that plasma proteomic signatures can reliably identify neurodegenerative processes across distinct populations, establishing a robust foundation for future clinical trials targeting pre-symptomatic ALS.