Machine learning empowered smartphone nanozyme sensing enables precise quantitative prediction of urinary uric acid

The increasing incidence of gout necessitates monitoring of uric acid (UA) levels for early risk assessment and clinical management. Although nanomaterials with oxidoreductase-mimicking properties demonstrate great promise in natural enzyme-free UA detection, accurate quantification in authentic samples remains a challenge due to common interference from coexisting reductants and matrix complexity. To overcome this limitation, here we integrate nanozyme catalysis-based smartphone sensing with machine learning to develop regression models that can effectively eliminate matrix interference, thereby achieving the accurate quantification of UA in real urine. A new peroxidase-mimetic nanozyme (Na2V6O16) was prepared to catalyze the H2O2-mediated oxidation of 3,3’,5,5’-tetramethylbenzidine (TMB) to blue oxTMB. When UA existed, it could suppress the chromogenic process due to its reducibility, leading to a smartphone-based assay exhibiting linear response to UA (0.05–0.4 mM) with a detection limit of 0.045 mM. Crucially, to eliminate the interference of redox species (ascorbic acid, L-cysteine, glutathione, etc.) coexisting in samples, machine learning algorithms optimized through parameter tuning were employed to correct background signals from actual specimens, significantly enhancing detection accuracy and reliability. Our work achieves the precise quantitative prediction of UA in complex urine matrices, providing a sensitive, reliable, and non-invasive approach for gout risk stratification and management.