Plasma protein changes linked to amyotrophic lateral sclerosis were detectable up to ten years before symptom onset, according to a recent study.
A study published in Nature Medicine evaluated plasma proteomics as a biomarker approach for amyotrophic lateral sclerosis (ALS). Investigators analyzed approximately 3,000 plasma proteins using the Olink Explore 3072 platform in samples from 231 patients with ALS and 384 controls, which included both healthy participants and those with other neurological conditions. Thirty-three proteins were identified as differentially abundant in ALS. Neurofilament light chain (NEFL) showed the greatest difference, consistent with prior research, while 31 additional proteins were associated with ALS. Results were replicated in an independent cohort and validated in external datasets.
Supervised machine learning was applied to proteomic data, incorporating clinical and demographic variables. A random forest model achieved an area under the curve of 98 percent when distinguishing ALS from controls, including patients with neurological disorders such as Parkinson’s disease, progressive supranuclear palsy, and myopathies. The model identified 20 predictive features, including 17 proteins, age, sex, and plasma collection tube type. When NEFL was excluded, predictive accuracy declined modestly, indicating that classification was supported by multiple proteins.
Pathway enrichment analysis of the 33 proteins demonstrated associations with skeletal muscle development, neuronal processes, and energy metabolism. The researchers also evaluated cerebrospinal fluid samples, where five proteins showed similar changes to those identified in plasma. Additional exploratory analyses in patients with C9orf72 repeat expansions identified eight proteins that were elevated in carriers compared with noncarriers, suggesting possible subgroup-specific profiles.
The machine learning model was further assessed in presymptomatic patients who later developed ALS. Among 110 such cases, the ALS risk score increased in relation to time to symptom onset, with changes observed up to 10 years before clinical diagnosis – earlier than previously observed. Ten proteins in the model, including NEFL, showed individual associations with time to onset.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
