Molecule-resolvable SERSome for metabolic profiling

Abstract

Multiplexed detection is a challenging yet essential task in analytical chemistry, especially for complex systems. Surface-enhanced Raman spectroscopy (SERS) is a promising analytical tool due to its molecular fingerprinting capability, sensitivity, low cost, and tractability. Considering the molecular profusion and diversity, SERSome, namely, spectral set, facilitates robust detection but is still challenged by spectral overlapping-induced uncertainty of molecular assignment and multiplexed quantification. Herein, we introduce molecule-resolvable (MORE) SERSome, identifying specific analytes contributing to the complex SERS spectra, which are then used in spectral decomposition for multiplexed analysis. Taking metabolic profiling for Alzheimer’s disease as a proof of concept, ten metabolites are screened in human serum. A deep-learning model enables accurate and rapid diagnosis, achieving an area under the receiver operating characteristic curve as high as 91.5%. Comparing with conventional methods, MORE SERSome presents a methodological advancement in multiplexed detection with strong potential for general applications and fundamental research in analytical chemistry.