Fourier transform infra red (FTIR) spectroscopy has drawn significant attention of scientific communities for discriminating normal and diseased conditions and establishing label free, non-invasive biomarkers for early detection of fatal diseases. Identifying subtypes of heterogeneous disease such as lung cancer improves knowledge about disease prognosis and prescribing proper therapeutics. FTIR based spectral diagnostic features to differentiate them are not yet reported. Present study attempted to classify lung cancer subtypes (24 squamous cell carcinoma or SQC, 20 adenocarcinoma or ACC and 20 small cell carcinoma or SCC) and 20 benign (BEN) cases by performing FTIR based computational analysis on serum samples and quantify changes in disease related biochemical components. Pre-processed spectral signatures were evaluated using ANOVA and principal component analysis- linear discriminant analysis (PCA-LDA). Significant differences in peak area corresponding to biochemical fingerprint (900–1800 cm-1) and high wavenumber (2800–3800 cm-1) regions were listed and these contributed to sufficient PCA-LDA classification of different disease types. Forward feature selection method was further applied to enlist significantly contributory spectral markers. Identified discriminating wavenumbers were assumed to be contributed by protein (1320, 1400, 1455, 1550, 1660 cm-1), lipid (2850, 2880, 2950, 3080 cm-1) and nucleic acids (1087, 1240 cm-1). Findings were validated by re-assessing the data using another in-house built binary classifier namely vector valued regularized kernel approximation (VVRKFA), to ensure classification potentiality of these spectral features. High classification accuracy (upto 90%) was achieved in distinguishing the subtypes. The study successfully demonstrated application of FTIR coupled with computational analytics for lung cancer subtype classification with higher precision and highlighted contributory spectral biomarkers, efficient for clinical decision making.

Fourier transform infra red spectroscopy based spectral biomarker identification: Applications in classifying lung cancer subtypes

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.