Determination of platinum-resistance of women with ovarian cancer by FTIR spectroscopy combined with multivariate analyses and machine learning methods

Abstract

Abstract Patients with high-grade ovarian cancer have a poor prognosis, thus effective treatment remains an unmet medical issue of high importance. Moreover, finding the reason for resistance to cisplatin is a crucial task for the improvement of anti-cancer drugs. In this study, we showed for the first time a chemical difference in a serum collected from platinum-resistance and platinum-sensitive women suffering from ovarian cancer using Fourier Transform InfraRed (FTIR) spectroscopy followed by a data analysis by Principal Component Analysis (PCA), Hierarchical Component Analysis (HCA) and 4 different machine learning algorithms. Obtained results showed a shift of PO2 -symmetric vibrations, amide III and amide II were observed on the FTIR spectrum of the serum collected from platinum-resistance women in comparison with the spectrum of the serum from platinum-sensitive women. Furthermore, PCA analysis clearly demonstrated the most important role of amide II and amide I in the differentiation of platinum-sensitive and platinum-resistance women. In addition, machine learning algorithms showed the important role of wavenumber at 1631 cm-1(amide I) and wavenumber at 2993 cm-1 (asymmetric stretching CH3 vibrations). The accuracy of the obtained results was above 92%. Summarizing, FTIR can be used in detection platinum-resistance phenomena.

Keywords

• FTIR
• Machine learning
• PCA
• Platinum-resistance women
• Ovarian cancer