Study of Organic Acids Profile in Human Urine by Capillary Zone Electrophoresis Aiming at the Diagnosis of COVID-19

Abstract

The coronavirus disease (COVID-19) pandemic highlighted the need for alternative diagnostic strategies based on accessible and non-invasive technologies. This proof-of-concept study evaluates capillary zone electrophoresis with ultraviolet detection as a data-acquisition platform for exploratory profiling of four urinary organic acids (tartrate, malate, lactate, and succinate) in samples collected from individuals tested by reverse transcription polymerase chain reaction (RT-PCR) for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Electropherogram data were processed using multiple machine learning algorithms to classify COVID-19-positive and -negative samples. The models included partial least squares discriminant analysis (PLS-DA), soft independent modeling of class analogy (SIMCA), decision tree, random forest, bagged trees, and stochastic gradient descent. Overall performance was moderate: PLS-DA showed limited discrimination, SIMCA achieved high sensitivity but low specificity, and tree-based models demonstrated balanced yet modest accuracy. The best results were obtained with the stochastic gradient descent classifier, reaching 70% accuracy and a Matthews correlation coefficient of 0.41. These findings indicate that urinary organic acid profiles provide only weak discriminatory power for COVID-19 classification. Nonetheless, the workflow demonstrates the feasibility of integrating capillary electrophoresis with machine learning modeling for exploratory, non-invasive metabolomics-based screening.

Keywords:
urine; capillary zone electrophoresis; target metabolomics; machine learning classification; COVID-19