Abstract
Currently, there is no objective biomarker to indicate disease progression and monitor therapeutic effects for amyotrophic lateral sclerosis (ALS). This study aimed to identify plasma biomarkers for ALS using a targeted metabolomics approach. Plasma levels of 185 metabolites in 36 ALS patients and 36 age- and sex-matched normal controls (NCs) were quantified using an assay combining liquid chromatography with tandem mass spectrometry and direct flow injection. Identified candidates were correlated with the scores of the revised ALS Functional Rating Scale (ALSFRS-r). Support vector machine (SVM) learning applied to selected metabolites was used to differentiate ALS and NC subjects. Forty-four metabolites differed significantly between ALS and NC subjects. Significant correlations with ALSFRS-r score were seen in 23 metabolites. Six of them showing potential to distinguish ALS from NC-asymmetric dimethylarginine (area under the curve (AUC): 0.829), creatinine (AUC: 0.803), methionine (AUC: 0.767), PC-acyl-alkyl C34:2 (AUC: 0.808), C34:2 (AUC: 0.763), and PC-acyl-acyl C42:2 (AUC: 0.751)-were selected for machine learning. The SVM algorithm using selected metabolites achieved good performance, with an AUC of 0.945. In conclusion, our findings indicate that a panel of metabolites were correlated with disease severity of ALS, which could be potential biomarkers for monitoring ALS progression and therapeutic effects.