The advent of data science and machine learning creates exciting real-world application-specific prospects for nanoparticle-based sensors (nanosensors) which operate based on the principles of surface-enhanced Raman scattering (SERS). At present, the translation of SERS-based nanosensors from laboratory to practice is hindered by the inability to fully comprehend convoluted analyte signals, interferences arising from complex real sample matrices and the inefficient and subjective nature of traditional manual visual inspections or multiple peak deconvolution strategies.
Yong Xiang’s research focuses on leveraging data science and machine learning to tackle these challenges faced by SERS-based nanosensors in the ultrasensitive identification and quantification of target analytes. Through strategic expansion of data dimensionalities, domain knowledge-driven feature engineering and feature selection, as well as data augmentation techniques, he investigates the ability to construct accurate predictive models even in the presence of potential matrix interferences. Importantly, he places significant emphasis on model interpretability and conducts feature importance analyses to elucidate new findings while ensuring that the high observed accuracies result from meaningful chemical interactions at the molecular level. His work highlights the key role of data science and machine learning in realizing numerous potential sensing and diagnostic applications spanning the food, biomedical and environmental industries.