Comprehensive Summary
This study investigated whether host-derived urinary proteins could serve as diagnostic biomarkers for Schistosoma haematobium infection in school-aged children. A total of 135 children, aged 7-15, were enrolled from two endemic communities of Zanzibar between 2021 and 2023. Urine samples were collected and analyzed with data-independent acquisition mass spectrometry which subsequently identified different expressed proteins (DEPs). Candidate biomarkers were identified through feature selection using the top 20 DEPs and 2 machine learning algorithms (LASSO and SVM-RFE) and then used to train six machine learning models (Bayesian, logistic regression, decision tree, SVM, random forest, and XGBoost). These methods were cross-validated with additional cohorts and standard egg microscopy served as the comparator. Proteomic profiling detected 823 common host proteins, of which 269 were differentially expressed between infected and healthy control children. Machine learning consistently highlighted five markers–SYNPO2, CD276, hnRNPM (which were down-regulated), and α2-macroglobulin (α2M), and LCAT (which were up-regulated)–as discriminative features. ELISA testing in two independent cohorts confirmed higher urinary α2M and LCAT in infected children. Combining the α2M and LCAT biomarkers improved performance throughout all the classifying models, with random forest and XGBoost achieving 100% accuracy in small validation sets.
Outcomes and Implications
Current diagnostics for Schistosoma haematobium rely on urine microscopy, which is insensitive for mild infections and difficult to implement in resource-limited settings. By identifying host urinary proteins as biomarkers, the study offers a potential route to simple, non-invasive tests that could improve detection in children, which is the group at highest risk. Clinically, combining biomarkers such as α2M and LCAT could form the basis of urine-based assays that avoid the variability of egg counts and enable earlier intervention. The work is still at the discovery and validation stage, and the authors note that larger, multi-site studies are needed before clinical translation. If further validated, this approach could be integrated into schistosomiasis control programs within the coming years as part of field-deployable diagnostic kits.