New Blood Test Shows Promise For Faster ALS Diagnosis

No more waiting months, or even over a year, for an ALS diagnosis?

A simple blood draw may eventually help doctors confirm ALS sooner. If validated in clinical settings, the new method could potentially transform care for patients facing one of medicine’s most devastating diseases. Researchers at the University of Michigan have developed a gene-based blood test that achieved nearly 90% accuracy in identifying amyotrophic lateral sclerosis patients, even when tested on completely independent patient groups the team had never analyzed before.

The breakthrough addresses a critical problem: ALS typically takes 5 to 15 months to diagnose after symptoms first appear, and some patients wait as long as 19 months for a definitive answer. During this window, the disease progresses. Patients cycle through specialists and undergo batteries of tests while grappling with diagnostic uncertainty.

Every month of delay matters because existing treatments work better when started earlier, and most clinical trials exclude patients whose disease has progressed too far. Patients manifest symptoms that resemble more common illnesses, leading to frequent misdiagnoses and errors before doctors identify the true culprit.

Testing Gene Activity in Blood Samples

The study, published in Nature Communications, analyzed blood samples from 422 ALS patients and 272 healthy controls at Michigan’s Pranger ALS Clinic between 2011 and 2021. Rather than measuring a single protein or biomarker, they used advanced RNA sequencing to track the activity of over 22,000 genes simultaneously, creating a comprehensive molecular portrait of what was happening inside each person’s blood cells.

The ALS patients reflected typical disease demographics: median age of 65 years, more males than females (58% versus 42%), and 87% reporting no known family history. They represented all major disease subtypes, with 26% experiencing bulbar onset affecting speech and swallowing first, 32% with cervical onset starting in arms and shoulders, and 39% with lumbar onset beginning in the legs. Their functional impairment levels varied, with a median score of 37 out of 48 on the standard ALS rating scale.

Comparing gene activity between ALS patients and controls revealed 3,640 genes showing significantly different expression: 1,999 more active in ALS patients and 1,641 less active. Many related to immune system function, aligning with mounting evidence that immune dysfunction contributes to ALS progression. Some of the most altered genes were involved in how cells recycle waste, move materials around, maintain muscle function, or carry out programmed cell death.

The team then trained seven different machine learning algorithms to recognize ALS patterns in blood gene expression data. XGBoost, the top performer, achieved an area under the curve of 0.91, a measure where 1.0 represents perfect accuracy and 0.5 represents random guessing.

Practical clinical application required narrowing thousands of genes down to manageable panels. The researchers created three streamlined versions containing 27, 29, and 30 genes, plus a combined 46-gene panel. These smaller sets are crucial because they could eventually translate into commercial diagnostic tests similar to existing cancer screening panels. On internal validation data, all panels performed exceptionally well. Accuracies ranged from 91.1% to 91.2%, sensitivities reached 93.2% to 94.2%, and specificities measured 86.0% to 87.9%.

Maintaining Accuracy in Independent Testing

The real validation came when the team tested their classifiers on a fully independent dataset from another research group: 86 ALS patients and 46 controls they’d never seen before. The combined 46-gene panel achieved an area under the curve of 0.894, showing the test’s accuracy held up in a completely separate group. Earlier attempts at blood-based ALS gene expression tests showed much lower accuracy in external validation, around 63% accuracy, 60% sensitivity, and 67% specificity, far below clinical utility thresholds.

The research extended beyond diagnosis into survival prediction. By combining gene expression data with standard clinical variables like age at symptom onset, sex, and initial symptom location, the team developed models that could separate patients into shorter, intermediate, and longer survival groups. When tested on the independent dataset, models incorporating gene features achieved wider separation between survival groups compared to using clinical variables alone. While not yet ready for clinical use, this approach may eventually give patients clearer expectations about their disease course.

Biological Pathways Disrupted in ALS

Examining which biological pathways showed the greatest disruption in ALS patients’ blood revealed disease-relevant processes. The ALS-related pathway itself appeared among the most disrupted, particularly after adjusting for differences in immune cell proportions between patients and controls. Other disrupted pathways included oxidative phosphorylation (the way cells produce energy), protein processing systems, and pathways shared with other neurodegenerative diseases like Parkinson’s and Huntington’s.

The researchers pushed their analysis further by identifying “core genes” altered not just in blood but also in nerve tissue directly affected by ALS. Comparing their blood gene expression data with published data from ALS patients‘ spinal cord tissue and laboratory-grown neurons carrying ALS-related genetic changes, they pinpointed genes central to the disease process across different tissue types.

These core genes became the foundation for drug discovery analysis using a massive database tracking how different compounds affect gene expression. The computational screening identified eight drug candidates potentially capable of reversing ALS-related gene expression changes. Some were already FDA-approved for other conditions, including ibrutinib (approved for certain blood cancers) and trifluoperazine (an antipsychotic medication). Others were experimental compounds with no previous ALS connection, opening entirely new research avenues. These results are only computer-based; they’ll need lab testing long before anyone considers them for treatment.

Advantages Over Single-Biomarker Approaches

ALS patients currently face median survival of just 2 to 4 years from diagnosis. Every month of diagnostic delay represents precious time lost for treatment initiation and trial enrollment. It also leaves patients less time to manage personal affairs. A blood test accelerating diagnosis by even several months could meaningfully improve both care quality and quality of life for thousands of patients annually.

Unlike measuring neurofilament light chain, a single biomarker that increases in many neurological conditions, not just ALS, gene expression panels may offer better specificity for ALS diagnosis, pending clinical validation.

The technology already exists to translate these findings into clinical practice. Similar gene expression panels like PAM50 are already FDA-cleared for breast cancer subtype classification, demonstrating the commercial and regulatory feasibility of this approach. An 18-gene panel for diagnosing high-grade prostate cancer from fluid samples recently entered clinical use, further validating the gene panel strategy.

The research team’s larger sample size, sophisticated RNA sequencing technology, and rigorous machine learning methods overcame the limitations that plagued earlier attempts at developing blood-based ALS gene expression tests. Their achievement in maintaining high accuracy across independent patient populations suggests the approach could be ready for next-phase testing against ALS-mimicking conditions and in presymptomatic carriers of ALS-causing genetic mutations.

---