---
title: "Integrated multi-omics detection of m6A‑SNP–related diagnostic biomarkers for ALS"
id: "frontiers-in-immunology-11-integrated-multi-omics-identification-of-m6a-snp-related-diagnostic-biomarkers"
canonical_url: "https://medichelpline.com/clinical-feed/frontiers-in-immunology-11-integrated-multi-omics-identification-of-m6a-snp-related-diagnostic-biomarkers"
content_type: "clinical_feed_article"
specialty: "Neurology"
source_name: "Frontiers in Immunology"
source_url: "https://www.frontiersin.org/articles/10.3389/fimmu.2026.1914869"
published_at: "2026-08-27T00:00:00.000Z"
evidence_level: "Journal Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Integrated multi-omics detection of m6A‑SNP–related diagnostic biomarkers for ALS
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/frontiers-in-immunology-11-integrated-multi-omics-identification-of-m6a-snp-related-diagnostic-biomarkers
- **Specialty:** [Neurology](https://medichelpline.com/clinical-feed/neurology.md)
- **Primary Source:** Frontiers in Immunology
- **Source URL:** [Original Journal Publication](https://www.frontiersin.org/articles/10.3389/fimmu.2026.1914869)
- **Published At:** 2026-08-27T00:00:00.000Z
- **Evidence Rating:** Journal Feed
## Executive GIST (TL;DR)
- This study integrates blood cis-eQTL data, RMVar m6A-SNP annotations, and ALS blood transcriptomes to identify **m6A-SNP**-related genes potentially useful as minimally invasive diagnostic biomarkers for **amyotrophic lateral sclerosis (ALS)**. - Data sources included eQTLGen cis-eQTL results (31,684 individuals) and two GEO whole-blood cohorts: GSE112676 (training; 741 samples: 233 ALS, 508 controls) and GSE112680 (validation; final 301 samples: 164 ALS, 137 controls). - After intersecting RMVar m6A-SNP annotations with cis-eQTL signals and differentially expressed genes, 109 candidate m6A-SNP–related genes were identified. - A combined machine-learning pipeline using random forest and LASSO reduced candidates to seven diagnostic markers: TMED5, OXR1, BRI3, FEM1C, SUZ12, EIF2AK4, and TJAP1. - A seven-gene diagnostic model and nomogram were trained on GSE112676 and tested unchanged on GSE112680; the combined model outperformed single markers in the training cohort and retained moderate discrimination in external validation (AUCs reported in supplementary data). - Immune deconvolution (CIBERSORT) revealed altered inferred immune-cell composition in ALS blood, including differences in monocytes, neutrophils, and T-cell subsets; correlations between candidate genes and immune-cell proportions were examined. - Predicted m6A modification sites near selected SNP loci were identified in silico using SRAMP; SNPs were annotated to nearby RBP-binding regions from available databases. - Exploratory clinical validation in a small peripheral-blood cohort (10 definite ALS patients, 10 healthy controls) showed significant upregulation of **FEM1C** and **SUZ12** at mRNA and protein levels and decreased global **m6A** levels in ALS samples. - The study emphasizes multi-omics integration for biomarker discovery but notes limitations: small exploratory clinical sample size, absence of complete nested cross-validation, unavailable clinical covariates in GEO datasets, and need for larger independent cohorts and further calibration before clinical use.
## Clinical Analysis & Structured Key Points
Frontiers | Integrated multi-omics identification of m6A-SNP-related diagnostic biomarkers in amyotrophic lateral sclerosis ORIGINAL RESEARCH article Front. Immunol. , 27 August 2026 Sec. Multiple Sclerosis and Neuroimmunology Volume 17 - 2026 | https://doi.org/10.3389/fimmu.2026.1914869 Published in Frontiers in Immunology Multiple Sclerosis and Neuroimmunology 7 impact factor 11.3 citescore Part of a Research Topic Emerging Molecular Biomarkers in Neurodegenerative Diseases Submission open 1316 views 2 articles Editor & Reviewers Edited by M B Matteo Becatti Reviewed by E B Eva Bagyinszky S Q Shizheng Qiu Outline Figures and Tables Figure 1 View in article Figure 2 View in article Figure 3 View in article Figure 4 View in article Figure 5 View in article Table 1 Clinical characteristics of ALS patients (n=10). View in article Table 2 Baseline demographic comparison of ALS patients (n=10) and healthy controls (n=10). View in article Table 3 Primer sequences used for quantitative real-time polymerase chain reactions. View in article Table 4 Antibodies used for Western blot analysis. View in article ORIGINAL RESEARCH article Front. Immunol. , 27 August 2026 Sec. Multiple Sclerosis and Neuroimmunology Volume 17 - 2026 | https://doi.org/10.3389/fimmu.2026.1914869 Integrated multi-omics identification of m6A-SNP-related diagnostic biomarkers in amyotrophic lateral sclerosis H W Hongfen Wang 1 † B S Bo Sun 2 † S W Shiya Wang 3 J P Jing Pan 4 R D Rongrong Du 1 X P Xinyuan Pang 1 J B Jiongming Bai 1 J H Jinming Han 5 * J F Jinli Feng 1 * 1. Senior Department of Neurology, Chinese People's Liberation Army General Hospital, Beijing, China 2. Department of Neurology, Beijing Chaoyang Hospital, Capital Medical University, Beijing, China 3. Outpatient Department of Zhantansi, Jingzhong Medical Zone, Chinese People's Liberation Army General Hospital, Beijing, China 4. Department of Respiratory and Critical Care Medicine, Second Medical Center, Chinese People's Liberation Army General Hospital, Beijing, China 5. Department of Neurology, Xuanwu Hospital, Beijing, China See more Article metrics View details Abstract Background: Amyotrophic lateral sclerosis (ALS) lacks reliable and minimally invasive biomarkers for early diagnosis. m6A-associated single-nucleotide polymorphisms (m6A-SNPs) may influence RNA methylation and gene expression, offering opportunities to identify clinically relevant diagnostic markers. Methods: We integrated eQTLGen cis-eQTL data, RMVar m6A-SNP annotations, and ALS transcriptomic datasets to identify m6A-SNP–related genes. Random Forest and LASSO regression were combined to screen robust diagnostic markers. A nomogram was constructed and validated using independent cohorts. Immune infiltration, predicted m6A modification sites, and potential RBP–SNP interactions were assessed. Peripheral blood samples from ALS patients were used for exploratory validation of gene expression and global m6A levels. Results: We identified 109 ALS-associated m6A-SNP-related genes with cis-eQTL signals and narrowed these to seven candidate diagnostic markers (TMED5, OXR1, BRI3, FEM1C, SUZ12, EIF2AK4, and TJAP1). The seven-gene model outperformed the individual markers in the training cohort and retained moderate discrimination in the independent validation cohort. ALS samples showed differences in inferred immune-cell composition, including monocytes, neutrophils, and T-cell subsets. The selected SNP loci were located near predicted m6A sites and annotated RBP-binding regions. Exploratory clinical validation showed significant upregulation of FEM1C and SUZ12 at both mRNA and protein levels, accompanied by reduced global m6A modification. Conclusions: Through multi-omics integration and exploratory clinical validation, this study identifies m6A-SNP-related candidate markers associated with ALS. The findings support further evaluation of m6A-related signatures for ALS discrimination and molecular characterization, while larger independent cohorts and additional calibration are required before clinical application. 1 Background Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by progressive degeneration of upper and lower motor neurons, resulting in muscle weakness, atrophy, and ultimately respiratory failure ( 1 ). Despite advances in understanding ALS genetics, early diagnosis remains challenging, and effective treatments remain limited ( 2 – 4 ). The etiology of ALS remains largely unknown, with approximately 10% of cases linked to specific genetic mutations (familial ALS). The remaining 90% of cases manifest as sporadic ALS ( 5 ). This type is associated with environmental factors, including exposure to toxins and infections ( 6 , 7 ). Although the majority of ALS cases are sporadic, genetic studies have revealed substantial molecular heterogeneity underlying ALS pathogenesis ( 7 ). More than 30 genes have been implicated in ALS susceptibility, with several major causative genes accounting for a significant proportion of familial and sporadic cases ( 8 ). Among these, C9orf72 repeat expansion, SOD1, TARDBP, and FUS are recognized as the most prevalent genetic contributors to ALS. The hexanucleotide repeat expansion in C9orf72 represents the most common genetic cause of ALS and frontotemporal dementia (FTD), accounting for a considerable proportion of familial ALS cases in European populations ( 9 ). Mutations in SOD1 cause ALS through mechanisms involving oxidative stress, mitochondrial dysfunction, and impaired protein homeostasis, whereas mutations in RNA-binding proteins such as TARDBP and FUS highlight the critical role of RNA metabolism dysregulation in ALS pathogenesis ( 10 – 12 ). In addition, other genes including TBK1, OPTN, VCP, UBQLN2, NEK1, and KIF5A have been identified as ALS-associated genes, further demonstrating the complex genetic architecture of this disease ( 13 , 14 ). Importantly, ALS and FTD are increasingly recognized as overlapping manifestations of a common neurodegenerative disease spectrum rather than completely distinct disorders. Approximately 10-15% of ALS patients develop clinical features of FTD, while a larger proportion exhibit subclinical cognitive and behavioral impairments associated with frontotemporal dysfunction ( 15 , 16 ). Conversely, a subset of FTD patients develop motor neuron degeneration resembling ALS, further supporting the existence of an ALS–FTD disease continuum ( 17 ). The pathological link between ALS and FTD is primarily characterized by abnormal aggregation and mislocalization of the RNA-binding protein TDP-43, which is observed in the majority of ALS cases and approximately half of FTD cases ( 18 , 19 ). Given that several ALS/FTD-associated genes, including C9orf72, TARDBP, FUS, and TBK1, are directly involved in RNA processing, RNA transport, and cellular homeostasis ( 20 , 21 ), dysregulation of post-transcriptional regulatory mechanisms may represent a convergent molecular mechanism underlying ALS-FTD spectrum disorders ( 1 ). Epigenetic regulation, including DNA methylation, histone modifications, and mRNA methylation, plays a growing role in the study of ALS pathogenesis ( 22 ). Among epitranscriptomic modifications, N6-methyladenosine (m6A) is a prevalent post-transcriptional modification in eukaryotic mRNA and influences mRNA stability, splicing, and translation ( 23 – 25 ). Increasing evidence suggests that m6A dysregulation is involved in neurodegenerative diseases, including Alzheimer’s disease (AD) and Parkinson’s disease (PD) ( 25 , 26 ), and may contribute to RNA- and protein-homeostasis pathways relevant to TDP-43 pathology. Thus, investigation of m6A modifications in ALS may support biomarker discovery and mechanistic studies. Previous integrated genomic analyses have identified m6A-SNP risk loci in other neurodegenerative diseases, including AD and PD ( 27 – 29 ). However, ALS-associated m6A-SNPs have not been systematically investigated, and their diagnostic relevance remains uncertain. 2 Methods 2.1 Experimental workflow of the ALS m6A-SNP integrated analysis study 2.2 Clinical sample collection The peripheral blood validation cohort included 10 patients with definite ALS and 10 healthy controls. Patient-level clinical characteristics of the ALS group are presented in Table 1 , while age and sex comparisons between ALS patients and controls are summarized in Table 2 . Patients were eligible for inclusion if they met the revised El Escorial criteria for definite ALS ( 30 ). This exploratory clinical validation study was approved by the Ethics Committee of the Chinese PLA General Hospital, Beijing, China (approval no. S2024-405-01), and informed consent was obtained from all participants or their legal guardians. Table 1 Patient ID Age Sex Site of onset Disease duration (months) ALSFRS-R Family history Mutation status Exclusion of other causes Patient 1 52 Male Upper limb 36 37 Negative Negative Yes Patient 2 47 Male Lower limb 12 42 Negative Negative Yes Patient 3 68 Male Upper limb 11 43 Negative Negative Yes Patient 4 60 Male Upper limb 24 41 Negative Negative Yes Patient 5 44 Female Upper limb 8 41 Negative Negative Yes Patient 6 63 Male Bulbar 12 45 Negative Negative Yes Patient 7 61 Female Bulbar 12 41 Negative Negative Yes Patient 8 53 Female Lower limb 12 40 Negative Negative Yes Patient 9 54 Male Upper limb 24 45 Negative Negative Yes Patient 10 53 Male Upper limb 7 47 Negative Negative Yes Clinical characteristics of ALS patients (n=10). Table 2 Variable ALS patients Healthy control P value Age (years), mean ± SD 55.5 ± 7.41 48.8 ± 14.76 0.216 Male sex, n (%) 7 (70.0%) 6 (60.0%) 0.639 Baseline demographic comparison of ALS patients (n=10) and healthy controls (n=10). 2.3 Data retrieval Blood expression quantitative trait locus (eQTL) data were retrieved from the eQTLGen consortium ( 31 ). This resource reports large-scale cis- and trans-eQTL analyses in blood, including data from 31,684 individuals. We used the available significant cis-eQTL results and allele-frequency information for downstream integration. Datasets GSE112676 and GSE112680 were selected as our bulk analysis cohorts, with GSE112676 serving as a training set and GSE112680 as a validation set. We downloaded expression profile data and clinical data from the Gene Expression Omnibus (GEO) database. GSE112676 included 741 whole-blood samples, including 233 ALS patients and 508 non-neurological disease controls. GSE112680 initially included 376 samples, including 164 ALS patients, 137 controls, and 75 ALS mimic cases. The ALS mimic samples were excluded before downstream analyses, resulting in 301 samples (164 ALS patients and 137 controls) in the final validation cohort. Both datasets were generated from PAXgene-stabilized whole blood samples. GSE112676 was based on the Illumina HumanHT-12 v3 Expression BeadChip platform (GPL6947), whereas GSE112680 was based on the Illumina HumanHT-12 v4 platform (GPL10558). The detailed characteristics of the GEO cohorts, including sample size, sample source, expression platform, and available clinical information, are summarized in Supplementary Table 1 . The GEO series matrix files had already undergone platform-specific preprocessing by the original data depositors, including normalization, log2 transformation, and quality-control procedures. The deposited metadata also documented technical procedures, including chip randomization and surrogate variable analysis (SVA), performed by the original investigators. These original preprocessing procedures were considered part of the GEO data processing and were distinguished from the downstream analyses performed in the present study. For downstream processing, empty probes and probes mapping to multiple genes were removed. When multiple probes corresponded to the same gene, the median expression value was used as the final gene-level expression value. The two GEO datasets were analyzed separately, with GSE112676 used as the training cohort and GSE112680 used as an independent external validation cohort. The datasets were not pooled for model construction, and batch-effect correction across datasets was therefore not performed in the present study. Available demographic and clinical characteristics, including sex distribution, age at disease onset, disease duration, and onset subtype, were summarized. Differences in sex distribution between ALS and control groups were evaluated using chi-square tests. Because age at onset, disease duration, and onset subtype were only available for ALS patients, direct age matching or adjustment across ALS and control groups could not be performed. Clinical variables including treatment information, disease duration at sampling, and blood-cell counts were unavailable in the GEO datasets and therefore were not included as covariates. Because direct blood-cell counts were unavailable, immune cell proportions were inferred from transcriptomic profiles using the CIBERSORT algorithm with the LM22 signature matrix. Correlations between inferred immune cell proportions and candidate marker genes were subsequently evaluated. 2.4 Analysis of ALS-associated N6-methyladenosine single nucleotide polymorphisms The RNA Modification-associated Variants (RMVar) database ( 32 ) was used as the source of RNA modification-associated variants, including m6A-related SNP annotations. RMVar integrates genetic variants with multiple RNA modification types and provides variant-level information for exploring potential effects on RNA modification. The RMVar database was used to cross-reference m6A-SNP data with eQTL data based on SNPs and genes. The resulting intersecting genes were further cross-checked against differentially expressed genes (DEGs) from our ALS bulk data to explore ALS-associated m6A-SNPs. Genes harboring both m6A-SNP annotations and cis-eQTL signals were considered m6A-SNP-related genes. These genes were subsequently intersected with ALS-related differentially expressed genes to identify candidate m6A-SNP-associated markers. 2.5 Gene ontology enrichment analysis To investigate the biological significance of the target genes, we performed functional enrichment analysis using the Metascape online platform. Gene ontology (GO) enrichment analysis describes the roles of genes in biological systems, including related biological processes (BPs), molecular functions (MFs), and cellular components (CCs). ALS-associated m6A-SNPs with a Benjamini-Hochberg (BH)-adjusted p -value 0.25 and an adjusted p-value <0.05 were considered differentially expressed. The results were visualized as heatmaps using the R package tidyheatmaps. 2.7 Machine learning screening for diagnostic marker genes To identify candidate diagnostic marker genes for ALS, we applied an integrated machine-learning strategy combining random forest and LASSO logistic regression. The outcome variable was binary coded as ALS = 1 and control = 0. Candidate genes were obtained by intersecting genes containing m6A-SNPs with cis-eQTL signals and ALS-associated differentially expressed genes. A total of 109 candidate genes were used for machine-learning screening. Key genes were screened using the random forest algorithm from the R randomForest package and the least absolute shrinkage and selection operator (LASSO) algorithm from the glmnet package. For random forest analysis, the mtry parameter was optimized using five-fold cross-validation, and a random forest model was constructed with 5,000 trees (ntree = 5000). Feature importance was evaluated according to MeanDecreaseGini values. The top 10 genes ranked by MeanDecreaseGini were selected as random forest-derived features. For LASSO analysis, penalized logistic regression was performed using the R glmnet package with family = “binomial” and alpha = 1. The optimal lambda value was selected using ten-fold cross-validation based on minimum binomial deviance. Genes with non-zero coefficients at the optimal lambda (lambda.min) were retained as LASSO-selected features. Genes identified by both random forest and LASSO were considered candidate marker genes, and their stability was further evaluated by bootstrap resampling. Feature stability was further evaluated using 100 stratified bootstrap resampling iterations. In each bootstrap iteration, ALS patients and controls were resampled while maintaining the original class proportion, and the complete feature-selection procedure, including random forest tuning, LASSO modeling, and feature intersection, was repeated independently. The selection frequency of each gene across the 100 iterations was calculated as the proportion of iterations in which the gene was selected. Genes with selection frequencies ≥req were considered stable features. Cross-validation was used for hyperparameter tuning within the training cohort: five-fold cross-validation was used to optimize the random-forest mtry parameter and ten-fold cross-validation was used to select the LASSO penalty parameter. The final feature-selection pipeline was developed using GSE112676, whereas GSE112680 was kept separate and was used only for external validation. Because a complete nested cross-validation estimate of the entire feature-selection pipeline was not generated, training-cohort performance is reported as apparent performance and the unchanged-coefficient external validation is treated as the primary assessment of generalizability. The seven-gene diagnostic model and nomogram were constructed in the training cohort using regression coefficients estimated from GSE112676. The derived coefficients were then applied unchanged to GSE112680 without coefficient re-estimation or model reconstruction. Receiver operating characteristic (ROC) curves were used to evaluate discrimination, with AUC values and 95% confidence intervals estimated using the DeLong method. Model calibration was assessed using the calibration intercept, calibration slope, Brier score, and Hosmer-Lemeshow goodness-of-fit test. Marker-specific and combined-model AUCs are summarized in Supplementary Table 2 , and calibration results are reported in Supplementary Table 3 . Recalibrated analyses in Supplementary Table 3 are presented as sensitivity analyses and do not replace the primary unchanged-coefficient external validation. 2.8 Immune microenvironment analysis The relative proportions of 22 immune cell types in the samples were estimated using the Type Identification by Relative Subset Estimation of RNA Transcripts (CIBERSORT) algorithm, implemented in the IOBR package. To assess whether immune cell infiltration profiles could distinguish patients with ALS from controls, principal component analysis (PCA) was performed using the FactoMineR package. Correlations between immune cell types were assessed with Spearman’s rank correlation in the psych package. A p -value of <0.05 was considered statistically significant. CIBERSORT results were visualized using the ggplot2 package, correlation heatmaps were generated with the corrplot package, and PCA results were visualized using the factoextra package. 2.9 Prediction of m6A modification sites We used the sequence-based RNA adenosine methylation site predictor (SRAMP) ( 33 ) to identify predicted m6A modification sites near the m6A-SNPs corresponding to the genes selected by machine learning. SRAMP is a sequence-based prediction tool for mammalian m6A sites; accordingly, these analyses were treated as in silico predictions rather than direct experimental evidence of methylation. 2.10 Interactions between m6A-single nucleotide polymorphisms and RNA-bindin
## Related Clinical Research

- [High-dose vitamin D (≥5,000 IU) linked to better cognition in older adults with MCI and sleep prob](https://medichelpline.com/clinical-feed/medical-news-today-1-high-dose-vitamin-d-may-boost-brain-heath-in-older-adults-at-risk-of-dementia.md)
- [Leisure physical activity lowers dementia risk while occupational activity may raise it, review fi](https://medichelpline.com/clinical-feed/medical-news-today-0-physical-activity-protects-brain-health-but-only-if-you-enjoy-it-study-says.md)
- [Metabolomics identifies lipid and amino acid signatures linked to disease severity in multiple scl](https://medichelpline.com/clinical-feed/biorxiv-3-metabolomics-reveals-lipid-and-amino-acid-signatures-of-disease-severity-in.md)
- [Climate change worsens health for people with chronic illness](https://medichelpline.com/clinical-feed/stat-news-1-opinion-climate-change-is-making-people-with-chronic-illness-even-sicker.md)
- [Resveratrol and the neuroinflammation axis in Alzheimer’s disease, Parkinson’s disease, multiple sclerosis, and cerebral ischemia](https://medichelpline.com/clinical-feed/frontiers-in-immunology-8-resveratrol-and-the-neuroinflammation-axis-in-alzheimer-s-disease-parkinson-s-disease-multiple-sclerosis-and-cerebral-ischemia.md)

## Navigation
- [← Back to Neurology Feed](https://medichelpline.com/clinical-feed/neurology.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.