Rapid identification and quantitative analysis of anthrax protective antigen based on surface-enhanced Raman scattering and convolutional neural networks

Abstract

A label-free detection method of the anthrax protective antigen (PA) based on surface-enhanced Raman scattering (SERS) was proposed. Au nanorods (AuNRs) substrates were prepared to realize the sensitive detection of PA. One-dimensional convolution neural network (1D-CNN) was used to process the Raman spectrum to achieve the qualitative and quantitative analysis of PA. The qualitative identification accuracy of PA under the interference of Human Serum Albumin (HSA) could reach 99.17%. In the quantitative prediction of PA concentration, the ability of CNN model (R2=0.856) was higher than that of the traditional partial least squares (PLS) method, which provides support for SERS quantitative analysis. Therefore, CNN could effectively identify and predict PA concentration with Raman spectrum, which would be helpful to expand the application of SERS technology in the field of the diagnosis of infectious disease.