Intelligent quantitative recognition of SARS-CoV-2 using machine learning-based ratiometric fluorescent paper sensors of metal-organic framework Al3+/Au NCs@ZIF-90

An advanced and highly sensitive analytical platform for SARS-CoV-2 is of crucial for public health. In this study, a machine learning-assisted platform that utilizes ratiometric fluorescent paper sensors based on the metal–organic framework Al³⁺/Au NCs@ZIF-90 was developed for precise and sensitive point-of-care testing (POCT) of SARS-CoV-2. This platform employs RdRp gene-induced hyperbranched rolling circle amplification (HRCA) to produce pyrophosphate (PPi) as a by-product, which triggers fluorescence quenching in ratiometric fluorescent paper sensors. Under ultraviolet (UV) excitation, the blue fluorescence emitted by ZIF-90 within Al³⁺/Au NCs@ZIF-90 serves as a reference signal, whereas the red fluorescence from Al³⁺/Au NCs acts as the analytical signal, with the fluorescence intensity being proportional to the PPi concentration. This approach not only ensures achieves high sensitivity but also exhibits a visible change of fluorescence color, achieving a limit of detection (LOD) of 0.3 pM specifically for SARS-CoV-2. By leveraging these distinctive fluorescence signals, the machine learning-assisted platform, which employs the Residual Neural Network (ResNet) algorithm, analyzes fluorescence images to discern SARS-CoV-2 RNA concentrations with an accuracy rate exceeding 99 %. The innovative platform integrates ratiometric fluorescent paper sensors with machine learning, offering a promising solution for point-of-care testing (POCT) of COVID-19 and potentially facilitating the early diagnosis of various diseases.