Comparative Analysis of Machine Learning Algorithms on Surface Enhanced Raman Spectra of Clinical Staphylococcus Species

Abstract

Raman spectroscopy (RS) is a widely used analytical technique based on the detection of molecular vibrations in a defined system, which generates Raman spectra that contain unique and highly resolved fingerprints of the system. However, the low intensity of normal Raman scattering effect greatly hinders its application. Recently, the newly emerged surface enhanced Raman spectroscopy (SERS) technique overcomes the problem by mixing metal nanoparticles such as gold and silver with samples, which greatly enhances signal intensity of Raman effects by orders of magnitudes when compared with regular RS. In clinical and research laboratories, SERS provides a great potential for fast, sensitive, label-free, and non-destructive microbial detection and identification with the assistance of appropriate machine learning (ML) algorithms. However, choosing an appropriate algorithm for a specific group of bacterial species remains challenging, because with the large volumes of data generated during SERS analysis not all algorithms could achieve a relatively high accuracy. In this study, we compared three unsupervised machine learning methods and 10 supervised machine learning methods, respectively, on 2,752 SERS spectra from 117 Staphylococcus strains belonging to nine clinically important Staphylococcus species in order to test the capacity of different machine learning methods for bacterial rapid differentiation and accurate prediction. According to the results, density-based spatial clustering of applications with noise (DBSCAN) showed the best clustering capacity (Rand index 0.9733) while convolutional neural network (CNN) topped all other supervised machine learning methods as the best model for predicting Staphylococcus species via SERS spectra (ACC 98.21%, AUC 99.93%). Taken together, this study shows that machine learning methods are capable of distinguishing closely related Staphylococcus species and therefore have great application potentials for bacterial pathogen diagnosis in clinical settings.

Discrimination between Carbapenem-Resistant and Carbapenem-Sensitive Klebsiella pneumoniae Strains through Computational Analysis of Surface-Enhanced Raman Spectra: a Pilot Study

“…For example, Wang et al ( 31 ) used CNN- and artificial neural network (ANN)-classified and predicted 18 Arcobacter species from clinical, environmental, and agri-food sources with an accuracy rate of 97.2%. In addition, Tang et al also successfully identified a set of clinically isolated Staphylococcus species via the combination of surface-enhanced Raman spectral fingerprinting and machine learning algorithms, which also confirmed the potential applicability of the SERS technology in clinical diagnostics ( 17 ). In terms of the differentiation of antibiotic resistance and sensitivity in bacterial strains, a variety of studies have addressed this question.…”

“…However, due to the complexity of Raman spectra, traditional linear analysis is not sufficient for the data-processing procedures, while machine learning (ML) algorithms are capable of extracting important features from the sophisticated SERS spectral data sets ( 15 , 16 ). Thus, SERS provides a great potential for fast and sensitive microbial detection and identification with the assistance of appropriate ML algorithms ( 17 ). At present, few studies have applied and compared machine learning methods in terms of SERS spectral analysis in order to distinguish between CSKP and CRKP strains.…”

Machine learning analysis of SERS fingerprinting for the rapid determination of Mycobacterium tuberculosis infection and drug resistance

“…Ho et al conducted the pioneering study in which the state-of-the-art CNN technique was applied to 60,000 SERS spectra for rapid identification of 30 common bacterial pathogens with the accuracy of 82% [33] . A series of studies then compared the prediction accuracies of various machine learning algorithms including deep learning algorithms in different types of bacterial pathogens [22] , [30] , [34] . In specificity, Tang et al performed the comparative analysis of ten machine learning algorithms on 2752 SERS spectra of nine Staphylococcus species, which confirmed that the convolutional neural network (CNN) algorithm was the best model with an accuracy of 98.21% [22] .…”

“…AgNPs synthesis followed the routine procedures that were previously reported by Tang et al with modifications [22] . Briefly, 33.72 mg of Silver Nitrate AgNO 3 (Sinopharm, Beijing, China) was dissolved into 200 mL ultra-pure water (deionized distilled water, ddH 2 O) with heating via a benchtop magnetic stirrer (ZNCL-BS230, Shi-Ji-Hua-Ke Pty.…”

Editorial: Computational Predictions, Dynamic Tracking, and Evolutionary Analysis of Antibiotic Resistance Through the Mining of Microbial Genomes and Metagenomic Data

“…In addition, bioinformatics analysis of the microbial genomes and metagenomic data would greatly facilitate our understanding of the molecular mechanisms, environmental transmissions, and dynamic changes of antibiotic resistance (De Abreu et al, 2021 ). Recently, many advanced bioinformatic methods, including the use of metagenomic next-generation sequencing (Berglund et al, 2019 ; De Abreu et al, 2021 ), machine learning (Liu et al, 2020 ; Anahtar et al, 2021 ), and Raman spectroscopy (RS) (Tang et al, 2021 ; Liu et al, 2022 ), have been proposed to predict ARGs and their mode of action. However, with steady accumulation of massively sequenced data and continuous antibiotic resistance emergence, novel and effective methodologies and tools for ARG prediction and antibiotic resistance profiling analysis and visualization are constantly needed.…”