Deep learning-enabled nanophotonic test leveraging poly-L lysine-tethered carbon quantum dots to assess the risk of cardiovascular disease

Abstract

The circulating cell-free mitochondrial DNA (ccf-mtDNA), in conjunction with circulating exosomes and NT-proBNP, represents a collective set of biomarkers crucial for the early identification and assessment of cardiovascular diseases (CVDs). In this context, our study explores an innovative quadruple-hybrid nano-photonic assay system engineered to identify all three distinct biomarkers. This cutting-edge assay comprises four sensing components: aminopyrene tethered carbon nanodots (CNDs) to efficiently detect circulating nucleic acids (ccf-NAs), followed by identification of ccf-mtDNA through PLL functionalized CNDs, and finally, targeting of NT-proBNP and circulating exosomes by graphene nanodots (GNDs-PLL) tethered with specific antibodies. The quadruple assay system demonstrated strong feasibility, high selectivity, and adaptability across various analytical platforms, with a detection limit of 0.369 fg and a quantification limit of 1.232 fg for ccf-mtDNA. The developed test overcomes the need for amplification and minimizes the detection time and assay-related complexities without disturbing the stability. The long short-term memory (LSTM) deep learning model accurately identified CVD and non-CVD samples with an overall accuracy rate of 99.86 % and validation loss of 0.25 % utilizing the sensor data and preliminary diagnostic data. The ability of the deep learning models to accurately discern CVD status suggests that integration of deep learning-based AI with biosensors further improves the precise diagnosis of CVD. Altogether, our observations highlight the promising potential of the developed AI-integrated biosensing system as a practical tool for comprehensive CVD risk assessment.