A blood protein panel may predict when ALS symptoms will begin

ALS usually becomes recognizable only after motor neurons have already been damaged and weakness has begun. Researchers now report that a pattern of proteins in blood may help estimate when symptoms are approaching in people with a high-risk genetic variant.

The result could improve the design of prevention trials by identifying participants near the transition to clinically apparent disease. It is an investigational biomarker panel, however, not a routine blood test that can currently predict ALS for the general public.

An unusual cohort made prediction possible

The work drew heavily on the long-running Pre-Symptomatic Familial ALS study, which follows people who carry pathogenic variants associated with ALS or frontotemporal dementia before they show clinical signs. The National Institutes of Health reported that the program has collected biological samples and clinical data for nearly two decades, creating a rare view of changes before and after symptom onset.

Researchers analyzed 516 plasma samples from 137 participants. The group included 33 people who later developed clinical manifestations, 35 people with ALS, 10 presymptomatic variant carriers who had not converted during observation and 59 controls. Most participants contributed samples at multiple visits, allowing the team to examine trajectories rather than a single snapshot.

Nineteen proteins formed the strongest timing panel

The team measured more than 5,000 proteins using a high-throughput proteomic platform. In the Nature Medicine study, 92 proteins changed before the transition to clinically manifest ALS or related frontotemporal disease. Machine-learning models then narrowed the information to a core panel of 19 proteins that best predicted transition across windows ranging from six months to five years.

The models produced cross-validated areas under the curve between 0.80 and 0.89 and estimated time to clinical onset with a mean absolute error of about 1.6 years. Several signals, including neurofilament light and proteins known as EDA2R and CA3, also appeared in UK Biobank data. That partial replication adds weight, but it does not amount to validation for individual clinical decisions.

An area under the curve summarizes how well a model separates cases from noncases across different thresholds; it is not the probability that a prediction for one person will be correct. Similarly, an average timing error of 1.6 years can hide wider misses in individual cases. The models were cross-validated within the available data, an important safeguard against overfitting, but prospective testing in a separate cohort remains the more demanding standard. For a rapidly progressive disease, even a one- or two-year timing difference can materially change when an experimental intervention begins.

Blood signals may capture different stages of disease

Neurofilament light rises as axons are injured, making it a useful marker close to symptom onset. Other proteins in the study shifted earlier and may reflect processes upstream of visible motor-neuron loss. A multi-protein panel can therefore carry information that a single marker misses, especially across people whose genetic forms of ALS progress at different speeds.

This distinction matters for prevention research. A trial launched too early may expose many participants to an experimental treatment years before they would otherwise develop symptoms. A trial launched too late may miss the best chance to preserve motor neurons. Better timing estimates could concentrate a study on people most likely to reach the clinical transition during the trial period.

The panel is tailored to a high-risk population

The strongest data came from carriers of known pathogenic variants, a small and distinctive subset of the population. The researchers note that such carriers are currently the only practical group for a long natural-history study beginning before symptoms. The panel therefore cannot be assumed to predict sporadic ALS, which accounts for most cases, or to screen people without a known elevated genetic risk.

The number of participants who transitioned was also modest. Repeated samples increase the amount of data, but they do not substitute for a large, independent validation cohort. Proteomic measurements can be sensitive to laboratory procedures, and predictive models may perform differently when transferred to new hospitals, ancestries or genetic variants.

The study also included transitions to clinically manifest ALS or frontotemporal dementia in its core longitudinal analysis because the two conditions can share genetic causes and biological features. That choice strengthens the analysis of a combined disease spectrum but makes careful labeling essential. A biomarker that forecasts a broader ALS-and-FTD transition is not automatically a precise forecast of the first motor symptom associated with classic ALS.

Prevention trials are the near-term destination

One reason for urgency is that gene-targeted treatment is moving earlier in the disease course. The ATLAS trial record describes a study of tofersen in presymptomatic carriers of certain SOD1 variants who show biomarker evidence of disease activity. Tofersen is approved for symptomatic SOD1-associated ALS, but whether earlier use can delay or prevent clinical onset remains under study.

A reliable timing panel could help match candidates to such trials and provide a clearer biological clock for researchers. That promise is substantial, yet the new work is best understood as a step in biomarker development. Clinical use will require standardized testing, prospective validation and evidence that acting on the prediction improves outcomes rather than merely forecasting them.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview