Rapid Surface-Enhanced Raman Scattering Imaging and Deep Learning for Highly Sensitive Discrimination of Amino Acids and Peptides

Abstract

Developing a highly sensitive and accurate method to discriminate between amino acids and peptides is vital for establishing future healthcare testing technologies, such as liquid biopsy. This study proposes a highly sensitive technique based on surface-enhanced Raman scattering (SERS), which combines chemically linking an analyte with gold nanoparticles and aggregating them to produce hotspots. Furthermore, by combining rapid SERS imaging with slit-scanning Raman microscopy and deep learning based on a convolutional neural network, 20 proteinogenic amino acids were successfully detected and distinguished with accuracies exceeding 95%. Also, out of 39 types of dipeptides that have Phe at either the amino terminal or the carboxyl terminal, 19 types were identified with high accuracy. Even for dipeptides with lower identification accuracy, it was confirmed that they were recognized as one of the dipeptides with high structural similarity, such as cyclic structures and branched amino acids. Moreover, pathophysiologically relevant sequence differences in β-amyloid peptides were accurately discriminated with a sensitivity of approximately 975 zeptomoles.