Panel of 19 Protein Plasmas May Predict Clinical Onset of ALS Years in Advance

An analysis of longitudinal plasma samples collected through 3 ALS natural history and biomarker studies—the Pre-Symptomatic Familial ALS Study (Pre-fALS; NCT00317616), Clinical Research in ALS Biomarker Study (CRiALS; NCT00136500), and CReATe Consortium Phenotype–Genotype–Biomarker Study (PGB1; NCT02327845)—identified a 19-protein panel that may help predict whether and approximately when individuals carrying amyotrophic lateral sclerosis (ALS)-associated pathogenic variants will develop clinically manifest disease. The findings could support patient selection and intervention timing in future ALS prevention trials.¹

Although ALS has traditionally been defined by the onset of progressive motor symptoms, accumulating evidence indicates that biological changes begin during a presymptomatic phase. Identifying these changes could enable investigational therapies to be administered before substantial neuronal injury and functional decline occur. However, reliable tools are still needed to determine which asymptomatic carriers of ALS-associated pathogenic variants are approaching phenoconversion, or the transition from presymptomatic disease to clinically manifest ALS.¹

Longitudinal Samples Capture Changes Before Symptom Onset

Researchers analyzed 516 serially collected plasma samples from 137 participants enrolled in longitudinal ALS biomarker and natural history studies. The discovery cohort included 33 individuals who phenoconverted to clinically manifest ALS and/or frontotemporal dementia, 35 patients with clinically manifest ALS, 10 presymptomatic carriers of ALS-associated pathogenic variants who had not phenoconverted, and 59 controls.¹

Using the Olink Explore high-throughput proteomics platform, researchers measured 5440 proteins, of which 5298 met quality-control criteria and were included in the analysis. They identified 137 proteins that were differentially expressed in patients with clinically manifest ALS compared with healthy controls, including 105 that were upregulated and 32 that were downregulated.¹

Pathway analyses showed that upregulated proteins were predominantly associated with skeletal muscle, whereas downregulated proteins were primarily linked to the extracellular matrix. Protein–protein interaction analyses also identified clusters related to skeletal muscle function, extracellular matrix biology, tumor necrosis factor–mediated signaling, regeneration, and neurofilament-associated processes.¹

Of the 137 differentially expressed proteins, 92 showed significant changes before phenoconversion. Concentrations of several proteins, including carbonic anhydrase 3 (CA3) and ectodysplasin A2 receptor (EDA2R), began increasing years before the onset of clinical manifestations. By contrast, neurofilament light chain (NfL), denoted as NEFL based on its encoding gene, increased sharply shortly before phenoconversion and continued to rise more gradually after symptom onset.¹

NfL is an established biomarker of axonal injury and has become increasingly important in ALS research and therapeutic development; however, because it reflects neuronal damage arising downstream of earlier disease mechanisms, NfL alone may not adequately capture the full presymptomatic course of ALS, particularly in carriers of pathogenic variants associated with more slowly progressive disease.^1,2

Multiprotein Model Outperforms NfL Alone

The researchers developed machine learning models to predict whether participants would phenoconvert within 6 months, 1 year, 2 years, 3 years, or 5 years of plasma collection. Depending on the prediction horizon, the optimized models incorporated between 11 and 34 proteins and achieved validation areas under the receiver operating characteristic curve (AUCs) ranging from 0.897 to 0.963. By comparison, models based on NEFL alone achieved AUCs ranging from 0.67 to 0.82.¹

The researchers subsequently identified a core 19-protein panel comprising NEFL, DUSP29, CALCA, NEB, NGRN, MYL11, NOS1, MYH1, TNNC1, DTNB, MEGF10, SYNM, APOA4, ART3, MYL3, EPHA1, EDA2R, TTN, and ACTN2. Collectively, these biomarkers reflect neuronal injury, skeletal muscle biology, and other biological processes that may contribute to ALS pathogenesis.¹

In cross-validated analyses, the 19-protein panel predicted phenoconversion across time horizons ranging from 6 months to 5 years, with AUCs of about 0.80 to 0.89. Depending on the prediction timeframe and selected cutoff, sensitivity and specificity each ranged from 76% to 95%.¹

The 19-protein panel generally outperformed NEFL alone, except when predicting phenoconversion within 6 months, for which the 2 approaches demonstrated comparable performance. The researchers suggested that the sharp increase in NfL shortly before phenoconversion may be sufficient for near-term prediction, whereas a multiprotein approach may provide greater predictive value over longer time horizons.¹

The 19-protein panel also estimated time to phenoconversion, with a mean absolute error of about 1.62 years and a correlation of 0.79 between predicted and observed timing. By comparison, NEFL alone yielded a mean absolute error of 2.37 years and a correlation of 0.48, with the 19-protein panel demonstrating significantly better model performance.¹

Findings Partially Replicated in UK Biobank

The researchers evaluated the findings in an independent replication cohort from the UK Biobank Pharma Proteomics Project. The cohort included 35,722 healthy controls, 38 presymptomatic carriers of SOD1 or C9orf72 pathogenic variants, 231 phenoconverters, and 22 individuals with clinically manifest ALS.¹

Despite key differences between the cohorts—including the use of cross-sectional rather than longitudinal proteomic data in the UK Biobank—several presymptomatic protein changes were replicated. These included increases in NEFL, EDA2R, CA3, DTNB, TNFRSF12A, HSPB6, and ITGB6. A reduced 15-protein panel comprising biomarkers available in the UK Biobank also outperformed NEFL alone in estimating time to phenoconversion.¹

Potential Implications for ALS Prevention Trials

The findings may have particular relevance to ALS prevention studies such as the phase 3 ATLAS trial (NCT04856982), which is evaluating whether tofersen (Qalsody; Biogen) can delay the onset of clinically manifest ALS in presymptomatic adults with SOD1 pathogenic variants and biomarker evidence of disease activity.³ In 2023, tofersen received accelerated FDA approval for the treatment of adults with SOD1-associated ALS based on reductions in plasma NfL, a surrogate biomarker considered reasonably likely to predict clinical benefit. Continued approval may depend on verification of clinical benefit in a confirmatory trial.⁴

A broader biomarker panel could help future trials identify carriers of ALS-associated pathogenic variants who are at increased short- or intermediate-term risk of phenoconversion, enroll participants at biologically meaningful disease stages, and assess whether treatment can delay or prevent clinically manifest ALS. However, the modest number of phenoconverters, differences between the discovery and replication cohorts, and focus on individuals with known pathogenic variants limit the findings’ immediate clinical applicability.¹

The models require validation in larger, independent, and genetically diverse cohorts before they can inform patient care or clinical trial enrollment. Regardless, the findings still support moving beyond reliance on a single marker of axonal injury toward a multiprotein approach that may better capture the broader biological changes preceding clinically manifest ALS.¹

REFERENCES

1. Ran, X., Wuu, J., Qin, Z.S. et al. Longitudinal plasma proteomics predict phenoconversion to clinically manifest ALS. Nat Med (2026). doi:10.1038/s41591-026-04528-x

2. Benatar M, Ostrow LW, Lewcock JW, et al. Biomarker Qualification for Neurofilament Light Chain in Amyotrophic Lateral Sclerosis: Theory and Practice. Ann Neurol. 2024;95(2):211-216. doi:10.1002/ana.26860

3. ClinicalTrials.gov. A study of BIIB067 (tofersen) initiated in clinically presymptomatic adults with a confirmed superoxide dismutase 1 mutation. ClinicalTrials.gov identifier: NCT04856982. Updated March 30, 2026. Accessed July 30, 2026. https://clinicaltrials.gov/study/NCT04856982

4. Qalsody. Prescribing information. Biogen Inc; 2023. https://www.biogencdn.com/us/pdfs/qalsody-prescribing-information.pdf