Colorimetric/chemiluminescent vertical flow immunoassay based on multifunctional cobalt–gold bimetallic nanoclusters for Brucella antibody detection and deep learning-assisted semiquantitative analysis

Abstract

A colorimetric/chemiluminescent vertical flow immunoassay (VFI) for Brucellosis was developed using glutathione-protected cobalt–gold nanoclusters conjugated with staphylococcal protein A (GSH@Co–AuNCs–SPA) as multifunctional catalytic immunoprobes. The nanocluster probe integrates antibody recognition with peroxidase-like catalytic activity, enabling three complementary signal readouts within a common immunorecognition architecture: direct colorimetric visualization, TMB–H₂O₂-amplified colorimetry, and luminol–H₂O₂ chemiluminescence. The prepared GSH@Co–AuNCs exhibited H₂O₂-mediated catalytic activity and sustained chemiluminescent emission, with a detectable visual signal persisting for more than 150 min, thereby providing a broad acquisition window for low-background imaging. Serial-dilution experiments showed experimentally determined signal-discrimination thresholds of 0.5 IU mL⁻¹ for direct colorimetry and 0.2 IU mL⁻¹ for both TMB-amplified colorimetry and chemiluminescence. For the chemiluminescent mode, analysis of 20 independent blank measurements yielded a statistical limit of detection of 0.134 IU mL⁻¹. Under the same analytical dilution-series conditions, a commercial colloidal-gold lateral-flow assay produced a visually distinguishable signal down to 20 IU mL⁻¹. To improve the consistency of image interpretation, shared-weight dual-branch ResNet-18 regression models were established for both TMB and chemiluminescent images. Grouped five-fold cross-validation yielded mean R² values of 0.8746 ± 0.0784 and 0.8907 ± 0.0588, with four-level grading accuracies of 96.25 ± 2.10% and 93.00 ± 6.28% for the TMB and chemiluminescent models, respectively. In a preliminary evaluation of 30 clinical serum samples, TMB-amplified colorimetry showed 95.0% sensitivity, 90.0% specificity, and 93.3% overall agreement with the serum agglutination test (SAT), whereas chemiluminescence showed 95.0%, 100%, and 96.7%, respectively. The proposed platform integrates catalytic signal amplification, complementary optical readouts, and regression-based semiquantitative interpretation within a single VFI system. This strategy provides a practical framework for improving weak-signal discrimination and standardized interpretation in rapid anti-Brucella antibody screening, while further validation in larger independent clinical cohorts remains necessary.