Identification of Reliable Biomarkers for ALS Through Machine Learning Approach

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by progressive motor neuron degeneration and limited diagnostic biomarkers. Identifying robust molecular biomarkers for ALS remains a major challenge due to disease heterogeneity and high-dimensional gene expression data. In this study, we developed a machine learning (ML) based pipeline integrating transcriptome data and feature selection to identify potential ALS biomarkers. RNA-Seq data of motor neuron disease patients and healthy controls were obtained from publicly available GEO datasets, followed by preprocessing was performed. We implemented two ensembled ML models such as eXtreme gradient boosting (XGBoost) and random forest (RF) algorithms under a five-fold stratified cross-validation framework to identify the differentially expressed genes. These models were evaluated using the performance metrics. We identified top 10 genes ranked by feature importance from the XGBoost and RF models. Notably, the DCN (Decorin) gene appears consistently in the top 10 features of both models, underscoring its stability and biological relevance. Both ML models exhibited excellent classification performance, with RF achieving 98.8% accuracy and XGBoost achieving 97.6% accuracy, alongside consistently high sensitivity, specificity, precision, and F1-score values. This work highlights the utility of transcriptomic data and ML in identifying key genes as biomarkers for diagnostic and therapeutic potential in ALS.