The research team, based at Houston Methodist Hospital’s MDA/ALSA ALS clinic, took advantage of a longitudinal serum repository collected between January 2018 and December 2022. From 352 patients with stored, immediately frozen samples, the investigators used a computer-generated random sampling process to select 100 patients for exploratory analysis. Each patient contributed between four and eleven blood draws, averaging six per person, spanning the full arc of disease from diagnosis to samples taken at or near end stage. One hundred age- and gender-matched healthy controls, typically spouses and friends of the patients, provided comparison samples. Only about five percent of the cohort carried known ALS mutations such as C9orf72 or a family history, making the group overwhelmingly representative of sporadic disease.

The three biomarkers were chosen for their biological coherence with what is known about ALS pathophysiology. 4-hydroxy-2-nonenal, or 4-HNE, is a toxic product of lipid peroxidation, the chemical assault on cell membranes that follows when dysfunctional mitochondria churn out reactive oxygen species. Lipopolysaccharide binding protein, or LBP, is synthesized and secreted by hepatocytes in the liver as part of the acute phase response, the innate, nonspecific first reaction of the body to immunological stress; it binds bacterial lipopolysaccharide and presents it to immune receptors such as CD14 and Toll-like receptor 4, amplifying inflammatory signaling. Neurofilament light chain, or NfL, is a structural protein released when axons degenerate, and it has already emerged as one of the most scrutinized fluid markers in neurodegeneration. All three were quantified by standard enzyme-linked immunosorbent assays, the kind of procedure routinely available in clinical diagnostic laboratories.

The cross-sectional findings immediately set ALS apart from other neurodegenerative diseases. Rapidly progressing patients, defined as those losing one or more points per month on the revised ALS Functional Rating Scale, showed significantly higher 4-HNE than both healthy controls and slowly progressing patients. LBP was elevated in rapid progressors compared with slow progressors and controls, and even slow progressors had higher LBP than controls. NfL followed a similar pattern, higher in rapid progressors than in slow ones and controls. Crucially, 4-HNE and LBP showed no such elevations in patients with Parkinson’s disease or mild cognitive impairment and Alzheimer’s disease, whereas NfL rose in those conditions as expected. This dissociation suggests that oxidative stress and systemic inflammation are not generic accompaniments of neurodegeneration but are specifically intensified in ALS.

The longitudinal data revealed a compelling divergence in how the three markers behave over time. At diagnosis, NfL showed the largest elevation relative to controls, roughly 3.3-fold, with LBP at about 1.6-fold and 4-HNE at a modest 1.3-fold. Yet from diagnosis to the last available sample, the picture inverted: 4-HNE climbed 177 percent and LBP rose 80 percent, while NfL increased only 27 percent, effectively plateauing. In other words, axonal injury as reflected by NfL is largely front-loaded into the early disease course, whereas lipid peroxidation and systemic inflammation escalate continuously as the disease advances. This temporal architecture helps explain why NfL did not correlate with ALSFRS-R scores in this cohort, since the functional scale kept declining steadily while NfL had already peaked.

The correlation analyses underscored the primacy of the inflammatory and oxidative markers. LBP at diagnosis showed the strongest negative correlation with ALSFRS-R score at diagnosis (r = −0.688) and the strongest correlation with survival from diagnosis to death (r = −0.773), along with the strongest link to progression rate (r = 0.668). 4-HNE tracked closely behind, correlating with survival at r = −0.661 and with progression rate at r = 0.589. NfL, by contrast, correlated only weakly with survival (r = −0.325) and progression rate (r = 0.274). The biomarkers also correlated with one another, moderately so between LBP and 4-HNE and between NfL and 4-HNE, painting a picture of interacting pathways in which oxidative damage, inflammation, and axonal breakdown feed one another. As expected clinically, older age at diagnosis and faster progression each correlated with shorter survival.

Site of disease onset added further biological texture. Among the 97 patients with either bulbar or limb onset, those with bulbar onset survived significantly shorter periods. Their mean 4-HNE levels were approximately 1.5-fold higher and LBP levels approximately 1.7-fold higher than in limb-onset patients, yet NfL levels did not differ between the two groups. This pattern implies that the worse prognosis of bulbar-onset ALS is driven not by greater axonal injury per se but by a more intense burden of oxidative stress and systemic inflammation, consistent with the idea that these pathways accelerate the pace of the disease rather than merely marking it.

The centerpiece of the study was a series of receiver operating characteristic analyses testing how well each biomarker, alone and in combination, could discriminate poor survival and rapid progression. Individually, 4-HNE and LBP performed well, with area under the curve values of 0.78 and 0.75 for poor survival and 0.83 and 0.74 for rapid progression, while NfL was more modest at 0.66 and 0.61. Pairwise combinations improved performance, and the three-marker panel achieved the best results: an AUC of 0.86 for poor survival and 0.84 for rapid progression, both highly significant. For third-quartile progression rates, the three-biomarker combination again reached an AUC of 0.86. Because post hoc ROC analyses risk overfitting, the team performed internal validation using bootstrap resampling with 2000 iterations. The optimism estimates were small, below 0.02, and the optimism-adjusted AUCs closely mirrored the apparent ones, with the three-marker survival model holding at an adjusted AUC of 0.85, indicating stable and robust discriminative performance.

The authors are candid about the limitations. The study was retrospective and single-center, drawing 100 patients from a larger repository, which may limit generalizability, though computer-generated random sampling minimized systematic selection bias and potential confounders including age, sex, genetic status, onset site, and disease duration were accounted for in the design. The ROC analyses lack independent prospective validation, which the bootstrap procedure only partially addresses. The study was explicitly framed as exploratory. Still, the practical appeal of the approach is considerable: unlike multi-protein panels discovered through untargeted proteomics and machine learning, these three biomarkers were selected on biological grounds, target three distinct and interacting pathways, and can be measured with standardized ELISA kits already familiar to clinical laboratories.

The findings also open a therapeutic horizon. Plasma NfL has already been validated as a pharmacodynamic marker through its reduction in association with tofersen-mediated slowing of disease progression in SOD1-mediated ALS. Whether LBP and 4-HNE can similarly serve as readouts of treatment response remains an open question, and the authors highlight prospective therapeutic studies as the necessary next step. If the escalating oxidative and inflammatory burden measured in this cohort truly reflects the engine of disease acceleration, then serial monitoring of these markers could one day tell clinicians whether a therapy is working months before functional scales register a difference. For a disease in which every point on the ALSFRS-R matters, a three-protein blood panel that discriminates fast from slow trajectories with an AUC near 0.86 is a meaningful step toward personalized prognosis and, ultimately, more responsive care.

Subject of Research: Longitudinal serum biomarkers for predicting ALS disease progression and survival

Article Title: Longitudinal Assessment of Biomarkers in ALS: Discriminative Biomarkers for Disease Progression and Survival

Article References: Beers, D. R., Lin, Y.-Y., Thonhoff, J. R., Thome, A. D., Faridar, A., Zhao, W., Wen, S., & Appel, S. H. (2026). Longitudinal Assessment of Biomarkers in ALS : Discriminative Biomarkers for Disease Progression and Survival. Annals of Clinical and Translational Neurology, 13(10), 2102-2113. https://doi.org/10.1002/acn3.70381

Image Credits: AI Generated

DOI: 10.1002/acn3.70381

Keywords: ALS, biomarkers, 4-HNE, lipopolysaccharide binding protein, neurofilament light chain, lipid peroxidation, systemic inflammation, axonal injury, survival, disease progression, ROC analysis, neurodegeneration

The clinical stakes of this distinction could hardly be higher. Alzheimer’s disease and frontotemporal dementia can look deceptively similar in the clinic, particularly in the earliest stages when symptoms are subtle and brain scans may still appear unremarkable. Yet the two conditions strike different neural circuits, progress along different trajectories, and respond to entirely different care strategies. Alzheimer’s disease typically begins in the medial temporal lobes and hippocampal networks, eroding memory before spreading outward, while frontotemporal dementia attacks the frontal and anterior temporal lobes, producing early changes in personality, behavior, and language. Misdiagnosis at this fork in the road can delay appropriate treatment, mislead families about what to expect, and complicate enrollment into clinical trials that increasingly depend on identifying the right patients at the right time.

Current gold-standard diagnostics rely on positron emission tomography imaging, cerebrospinal fluid analysis, and, more recently, plasma biomarkers, all of which carry substantial cost, limited availability, or patient burden. Electroencephalography, by contrast, is cheap, painless, widely available in hospitals around the world, and capable of capturing the brain’s oscillatory dynamics with millisecond resolution. The idea that dementia leaves fingerprints in these rhythms is not new; decades of research have documented slowing of background activity and disrupted functional connectivity in affected brains. What has been missing is a rigorous, interpretable way to translate those signals into a reliable differential diagnosis, and that is precisely the gap the Turkish team set out to close.

The study drew on a publicly available dataset of routine scalp EEG recordings from patients with Alzheimer’s disease, patients with frontotemporal dementia, and healthy individuals, a collection originally curated and described by Miltiadous and colleagues. Rather than feeding raw waveforms into a deep neural network, the researchers took a feature-engineering approach, extracting a rich library of measurements from each recording. These included time-domain statistics that describe the amplitude and distribution of the voltage signal, as well as frequency-domain features that quantify how the brain’s power is distributed across the canonical EEG bands, from the slow delta and theta rhythms to the faster alpha and beta oscillations. Such features encode well-established neurophysiological phenomena, including the posterior alpha slowing characteristic of Alzheimer’s pathology and the frontal-predominant changes associated with frontotemporal degeneration.

Crucially, the team did not rely on brain data alone. They combined the EEG-derived features with demographic and cognitive information, creating a genuinely multimodal input space that mirrors how a clinician actually reasons, weighing test results alongside age, sex, and cognitive scores. To prevent this expanded feature set from overwhelming their relatively small sample, the researchers applied the Fast Correlation-Based Filter, a feature selection method introduced by Yu and Liu that prunes redundant and weakly informative variables by evaluating correlations between features and their relationship to the class label. The result was a compact, information-dense feature vector that preserved the most diagnostic signals while discarding noise, a critical step when the number of measurements threatens to exceed the number of subjects.

With the feature space curated, the authors benchmarked an arsenal of classical machine learning models: XGBoost, Random Forest, AdaBoost, artificial neural networks, and Gradient Boosting. Each was evaluated under leave-one-out cross-validation, a demanding scheme in which the model is trained on all but one subject and then asked to classify the held-out individual, repeated across every participant. In small-sample biomedical research, this procedure is widely regarded as the most honest way to estimate generalization performance, though it is computationally expensive and leaves no room for the model to hide from difficult cases. XGBoost emerged as the clear winner, an outcome consistent with the algorithm’s track record on structured tabular data, where its regularized boosting framework excels at capturing nonlinear interactions among heterogeneous features without requiring massive datasets.

What elevates the study beyond a leaderboard exercise is its commitment to explainability. Black-box predictions have long been a barrier to clinical adoption, and regulators and physicians alike are increasingly unwilling to act on outputs they cannot interrogate. Seçkin and Soygazi addressed this by applying SHAP, or SHapley Additive exPlanations, a technique rooted in cooperative game theory that assigns each feature a consistent, individualized contribution to every prediction. Rather than asking simply whether the model was right, the SHAP analysis reveals which specific measurements pushed each classification toward Alzheimer’s disease or frontotemporal dementia, and by how much. This transforms the model from an oracle into a collaborator, allowing neurologists to inspect the evidence behind a diagnosis and judge whether it aligns with the clinical picture.

The interpretability layer also serves a scientific purpose. By aggregating SHAP values across the cohort, researchers can identify which neurophysiological and clinical variables carry the greatest discriminative weight, generating hypotheses about the underlying pathophysiology. If, for example, particular spectral features from frontal electrode sites dominate the separation between the two dementias, that pattern itself becomes a biomarker candidate worth validating in prospective studies. The approach situates the work within a rapidly growing literature on explainable artificial intelligence in dementia research, including recent scoping reviews of xAI in dementia detection and prior explainable deep learning frameworks for EEG-based diagnosis, while distinguishing itself through its explicit multimodal design and its head-to-head comparison of interpretable tree ensembles.

The authors are careful to frame the results as promising rather than definitive. The dataset, while valuable, is modest in size, and the leave-one-out design, though rigorous, cannot substitute for external validation on independent cohorts recruited in different clinical settings with different equipment and preprocessing pipelines. Routine EEG recordings also carry artifacts and variability that a research-grade dataset may not fully capture. The authors note that no new data were collected for the study, meaning that ethical approval was not required, and they declare no conflicts of interest. The underlying dataset remains available to other researchers upon reasonable request, an openness that should accelerate replication efforts.

Even with those caveats, the implications are compelling. A diagnostic aid built on a machine that already exists in most neurology departments, requiring no contrast agents, no lumbar punctures, and no radioactive tracers, could dramatically widen access to early differential diagnosis, particularly in regions where advanced imaging is scarce. Combined with the transparency afforded by SHAP analysis, such a tool could earn the trust of clinicians rather than merely impressing them with accuracy statistics. The next steps, external validation, prospective testing, and integration into clinical workflows, are demanding, but the study demonstrates that the humble EEG, married to carefully engineered features and explainable machine learning, can do more than record brain activity. It can help answer one of dementia medicine’s most consequential questions with nearly ninety percent confidence, and it can show its work while doing so.

Subject of Research: Explainable multimodal machine learning using EEG and clinical data for early differential diagnosis of Alzheimer's disease and frontotemporal dementia

Article Title: Explainable multimodal learning for early differential diagnosis oF Alzheimer’S disease and frontotemporal dementia

Article References: Seçkin, A. Ç., & Soygazi, F. (2026). Explainable multimodal learning for early differential diagnosis oF Alzheimer’S disease and frontotemporal dementia. Neural Computing and Applications, 38(17), Article 715. https://doi.org/10.1007/s00521-026-12444-x

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12444-x

Keywords: Alzheimer's disease, frontotemporal dementia, EEG, machine learning, explainable AI, XGBoost, SHAP, feature selection, dementia diagnosis, neurodegenerative diseases, clinical decision support, biomarkers