Abstract

Metabolomic detection enables a systems-level understanding of biological processes, while many emerging demands remain unmet. Surface-enhanced Raman spectroscopy (SERS) has recently evolved into a promising platform for metabolic detection yet not reaching true metabolomics. This Mini-Review is motivated by recent advances in understanding molecule–nanomaterial interactions aimed at addressing the related limitations. We first outline the fundamental principles enabling SERS-based metabolomic detection, including specificity, sensitivity, near-field compatibility with small metabolites, and nondestructiveness. Gaps between current SERS techniques and true metabolomics are delineated, and the key technical advances to overcome these challenges are also highlighted, including digital SERS, SERSome, molecule-resolvable SERSome, probe-functionalized nanomaterials, and artificial intelligence-assisted analysis. These developments have enabled SERS across multiple analytical paradigms, spanning targeted detection, phenotypic profiling, and emerging metabolomics. At last, we discuss the future challenges in hopes of advancing SERS from a sensing-oriented technique to a true metabolomic platform, ultimately facilitating the decoding of biological systems.

Figure 1. Mechanistic basis enabling SERS for metabolomic detection. (a) The single-molecule detectability and molecular specificity of SERS enables ultrasensitive identification of metabolite species via the vibrational fingerprints. (b) Metabolites possess sizes comparable to electromagnetic hotspots, in contrast to larger macromolecules, rendering multiplexed metabolite detection more tractable. (c) The noninvasive optical nature of SERS permits dynamic and potentially real-time metabolic monitoring.

Figure 2. Recent technical advancements overcoming the core challenges toward SERS-based metabolomic detection. (a) Accurate quantification of trace analytes is generally challenged by single-molecule intensity fluctuations, while the digital SERS protocol converts analog spectral responses into binary events to improve quantitative reproducibility. By calculating the ratio of positive voxels (RPV) within a spectral set, quantification error follows Poisson statistics and decreases with increasing event counts. (b) Label-free SERS detection of complex biosamples is challenged by the varied signatures at different acquisitions and ambiguous interpretation for the molecular origin of the SERS signatures. A SERS spectral set (SERSome) can be constructed via pointwise mapping within biosample–colloid mixtures to enhance profiling robustness. The minimal number of spectra required for reliable profiling can be estimated using statistical metrics such as Pearson’s correlation coefficient (PCC) and Wasserstein distance (WD). Multiplexed metabolite detection is achieved through spectral deconvolution into contributions from individual metabolites. (c) Though certain surface chemistry on the SERS nanomaterials limits the scope of detectable molecules, specific surface functionalization of SERS nanomaterials with diverse probe molecules expands metabolite coverage by introducing selective enrichment. (d) The complexity of molecular composition in the biosample results in a variable background matrix effect and entangled molecular signature. Artificial intelligence techniques boost the analysis of SERS-based metabolomic data, including using transfer learning strategies to facilitate quantification across distinct background matrices when comprehensive labeled data sets are unavailable and using machine learning or deep learning methods coupled with model explanation for diagnosis and biomarker identification.

https://doi.org/10.1021/acs.nanolett.6c02066