Label-free detection of high-risk human papillomaviruses infection using Raman spectroscopy and multivariate analysis. (22nd September 2020)

• Record Type: Journal Article
• Authors: Zheng, Xiangxiang; Wang, Jing; Yin, Longfei; Luo, Bin; Lv, Xiaoyi; Wu, Guohua
• Is Part Of: Laser physics letters. Volume 17:Number 11(2020:Nov.)
• Publication Date: 2020-09-22
• Subjects: Raman spectroscopy -- high-risk human papillomaviruses -- cervical exfoliated cell specimens -- principal component analysis -- linear discriminant analysis; Lasers -- Periodicals; Physics -- Periodicals; 621.366
• DOI: 10.1088/1612-202X/abafbf
• Languages: English
• ISSNs: 1612-2011

Abstract

Abstract: Persistent infection with high-risk human papillomavirus (HR-HPV) is considered as the leading cause of pre-cervical cancer and cancer. Currently, HPV testing is expensive, time-consuming, and requires experienced technicians. Here, we report a method for label-free detection of HR-HPV by recording and analyzing Raman spectroscopy (RS) of cervical exfoliated cell specimens. A total of 315 Raman spectra (165 normal, 150 HR-HPV) were recorded from 63 subjects (33 normal samples, 30 HR-HPV positive patients). Mean Raman spectra differed significantly between normal and HR-HPV positive groups. The differences showed an increase in the relative contents of hemoglobin and tryptophan, and new peaks appeared due to cell-level infection in HR-HPV positive group. Furthermore, a principal component analysis-linear discriminant analysis diagnostic model was developed and applied on the Raman spectra of HR-HPV positive patients as well as normal samples, and satisfactory classification results were obtained. Raman diagnostic performance was evaluated through the leave-one-spectrum-out cross-validation (LOSOCV) or leave-one-patient-out cross-validation (LOPOCV) method. The diagnostic accuracy, sensitivity and specificity of LOSOCV and LOPOCV were 99.4%, 99.4%, 99.3% and 98.4%, 100%, 96.7%, respectively. Interestingly, the use of LOSOCV or LOPOCV method yielded similar classification results. The results showed that the use of multiple data points from the same sample would not lead to biases necessarily impacting the reliability of the classification models. Furthermore, our exploratory work demonstrated that RS combined with multivariate analysis might provide a new method for clinical detection of patients with HR-HPV infection.