Dual-Mode SERS Lateral Flow Aptamer Assay with Machine Learning-Driven Highly Sensitive Interferon-γ Detection

Interferon-γ (IFN-γ), a key pro-inflammatory cytokine, is widely recognized as a critical biomarker for diagnosing and monitoring various immune-related conditions. However, its typically low concentrations in biological fluids─at the picogram-per-milliliter (pg/mL) level─necessitate ultrasensitive detection strategies for early clinical intervention. Here, we report a dual-mode surface-enhanced Raman scattering (SERS) lateral flow aptamer assay that employs a competitive binding mechanism between IFN-γ and its complementary DNA for aptamer recognition. This platform combines visual readout with quantitative SERS detection, enabling accurate measurement over a wide dynamic range (5-2000 pg/mL) with a limit of detection of 2.23 pg/mL. Clinical validation using human serum samples confirmed the assay's ability to distinguish IFN-γ concentration tiers─negative, low, and medium/high─with high diagnostic accuracy, supporting its potential for point-of-care applications. To enhance interpretability and classification performance, the system was integrated with machine learning algorithms, including multinomial logistic regression (MLR), multilayer perceptron, and random forest. Among these, the MLR model achieved the best performance, with an overall accuracy of 94.12% and a macro-average area under the ROC curve of 1.00. It further demonstrated group-specific sensitivities and specificities of 100% for the negative group, 83.33%/100% for the low group, and 100%/90.91% for the medium/high group. This dual-mode, machine learning-assisted biosensing platform offers a robust and practical solution for ultrasensitive cytokine detection, bridging the gap between analytical performance and clinical applicability in precision diagnostics.