Peripheral immune patterns enable robust cross-platform prediction of ALS onset and progression.

Amyotrophic lateral sclerosis (ALS) progression rates vary dramatically between patients, yet the basis of this heterogeneity remains elusive, with no prognostic biomarkers existing to guide clinical decisions or stratify patients for therapeutic trials. Here, we identify a network of coordinated immune cell types, which exhibit differential disruption across progression groups. Using mass cytometry (CyTOF) to profile 2.2 million immune cells from 35 ALS patients stratified by progression rate and 9 healthy controls, we find that the extent of immune dysfunction cannot be reflected by examining differences in individual cell type frequencies. In contrast, analyses of correlation patterns between cell types revealed distinct immune organization patterns, where coordination complexity varied with disease progression. Across all progression groups, we observed striking immune reorganization in natural killer (NK) cells and a major shift from B cell/basophil coordination hubs in healthy controls to neutrophil/T cell-dominated patterns in ALS. Having established coordinated immune patterns, we developed machine learning models to further improve our ability to stratify between disease and non-disease cohorts, achieving superior performance compared to models using cell frequencies alone. Central and effector memory (CM/EM) CD4+ T cell interactions emerged as top discriminative features for disease status, while plasmacytoid dendritic cell (pDC) relationships, especially their ratio with regulatory T cells (T-regs), distinguished progression rates, supporting T-reg-based therapeutic approaches. These findings reframe ALS as a disease of immune coordination breakdown, pointing towards cell-type specific therapeutics and biomarkers that may extend beyond ALS to other neurodegenerative diseases characterized by immune dysfunction.