AI and Neurological Rare & Neuroimmune Diseases ── ALS ── 2026-10-12
ALS: treatment, trials and AI
- Abstract
Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder characterized by the progressive loss of motor neurons. It is caused by multiple factors, and understanding its molecular pathogenesis remains challenging, particularly in the early stages for identifying diagnostic biomarkers and effective treatments. In this study, we aimed to identify novel biomarkers in ALS and elucidate their molecular mechanisms using transcriptomics and artificial intelligence approaches. We considered two RNA-seq datasets of ALS, namely GSE277709 and GSE234297, obtained from the NCBI GEO database. Initially, the data were preprocessed, which included quality control and filtering. Common genes (12,281 genes) between the datasets were identified and merged into a unified dataset comprising 216 samples. Batch effects were corrected using ComBat-seq and validated through principal component analysis and ANOVA-based $$R^2$$ analysis. We applied machine learning (ML) models including logistic regression (LR), support vector machine (SVM), random forest (RF), and eXtreme Gradient Boosting (XGBoost) to identify differentially expressed genes (DEGs). The identified biomarkers were further validated using DESeq2 log $$_2$$ fold change with a threshold of 0.6 and p value $$< 0.05$$ . Finally, functional enrichment analysis was performed on the identified DEGs using Enrichr tools. Our results demonstrated that batch correction significantly reduced batch-driven separation, with principal component variance decreasing from 70.2 to 19.9% before and after correction, respectively. Furthermore, ANOVA-based $$R^2$$ analysis confirmed an 86.65% reduction in batch-associated variance. We identified a total of 29 coding genes and 38 non-coding genes (67 genes) between DESeq2 and ML models. However, few genes are common between ML models and we received a total of 41 genes (18 coding genes and 22 non-coding genes, and gene symbols were not available for one Ensembl IDs). Our ML models such as LR, SVM, RF, and XGBoost, achieved balanced accuracies of 77.46%, 80%, 77%, and 75.82%, respectively. The enrichment analysis revealed that 10 DEGs were associated with 34 ALS-related biological pathways. Notably, CCND1 and MYL9 was found to be involved in GO, KEGG and Reactome pathways. Overall, our results demonstrate that computational models combining transcriptomics and artificial intelligence can effectively support biomarker identification and pathway discovery in ALS.
Journal IF-equivalent: 3.1 (OpenAlex 2-year mean citedness, value as of 2026-10-11, retrieved 2026-10-12; not the official Clarivate IF)Reference: Yadav R, Sriram Kumar P, Pragya P, Ronickom JFA. Computational models for biomarker identification in amyotrophic lateral sclerosis using transcriptomics and artificial intelligence. Discov Artif Intell. 2026 Oct 10;6(1):1421. doi:10.1007/s44163-026-02415-5.Checked: Abstract only