ALS is a fatal neurodegenerative disease characterized by the progressive loss of motor neurons, leading to muscle weakness, paralysis, and eventual respiratory failure. Because diagnosis is currently based on clinical symptoms and electrophysiological findings only appearing after significant neuronal damage has already occurred, there is a strong need for accessible biomarkers that reflect disease-specific molecular changes and can support earlier and more accurate diagnosis. Blood-based biomarkers are particularly attractive due to their practicality, but identifying robust transcriptomic signatures has been challenging, in part because ALS is biologically heterogeneous and driven by complex molecular interactions that are not well captured by conventional linear analyses.

To address this challenge, the research team applied a non-linear, data-driven approach to analyze gene expression profiles from peripheral blood mononuclear cells. Rather than focusing on individual genes, the method evaluates gene combinations and measures how strongly their collective expression patterns differ between patients and healthy controls. Using large publicly available blood transcriptome datasets, the analysis systematically searched for gene combinations that maximized group separation while minimizing redundancy among genes.

This approach identified a three-gene signature composed of PRKAR1A, QPCT, and TMEM71 that distinguished individuals with ALS from healthy subjects with high accuracy. The combined expression pattern of these genes achieved strong classification performance in public datasets and was further validated in independently collected blood samples analyzed in the laboratory. Importantly, although each gene showed only modest individual changes, their combined behavior provided clear discriminatory power, highlighting the value of capturing non-linear interactions rather than relying on single-gene markers.

To examine whether this blood-derived gene signature reflects biologically meaningful disease processes, the researchers extended their analysis to motor neurons generated from human iPS cells. While expression differences of the three genes were subtle at the single-gene level, their combined expression pattern also enabled classification of ALS and control motor neurons. Functional experiments further demonstrated a direct link between the identified genes and ALS-related pathology. Reducing the expression of each gene in human motor neurons led to increased levels of TDP 43, a protein central to ALS pathology. Notably, suppression of PRKAR1A caused abnormal redistribution and phosphorylation of TDP 43 in ALS motor neurons, accompanied by structural degeneration of neuronal projections, changes that mirror key features of the disease.

These findings suggest that the identified gene combination is not merely a diagnostic signal but is connected to molecular mechanisms that contribute to neuronal dysfunction in ALS. Although the signature showed limited ability to distinguish ALS from other neurological disorders, its performance supports its potential use as a supportive blood-based biomarker rather than a standalone diagnostic tool.

By integrating non-linear computational analysis with experimental validation in human motor neurons, this study demonstrates a pathway from unbiased data exploration to mechanistic insight. The work highlights the potential of advanced analytical frameworks to uncover hidden molecular patterns in accessible tissues and provides a new perspective on how blood-based transcriptomic data can inform both biomarker development and understanding of ALS pathogenesis.

• doi: 10.1093/biomethods/bpag023