Spectral cross-correlation as a supervised approach for the analysis of complex Raman datasets: the case of nanoparticles in biological cells

Abstract

Spectral cross-correlation is introduced as a methodology to identify the presence and subcellular distribution of nanoparticles in cells. Raman microscopy is employed to spectroscopically image biological cells previously exposed to polystyrene nanoparticles, as a model for the study of nano–bio interactions. The limitations of previously deployed strategies of K-means clustering analysis and principal component analysis are discussed and a novel methodology of spectral cross-correlation analysis is introduced and compared with the performance of classical least squares analysis, in both unsupervised and supervised modes. The previous study demonstrated the feasibility of using Raman spectroscopy to map cells and identify polystyrene nanoparticles in a lipid rich environment, which is suggestive of the membrane rich endoplasmic reticulum. However, short comings in identification of all nanoparticle signatures in the cell using K-means clustering are apparent, as highlighted by principal component analysis of the identified clusters which demonstrates that K-means clustering does not identify all regions where spectral signatures of the nanoparticles are evident. Thus, two more sophisticated analytical approaches to the extraction of the nanoparticle signatures from the Raman spectral datasets, namely classical least squares analysis and cross-correlation analysis, were employed and are demonstrated to improve the identification of spectroscopic signatures characteristic of polystyrene nanoparticles in a cellular environment. Additionally, to investigate the local biochemical environment in which the nanoparticles are trafficked, a pure spectrum of 3-sn-phosphatidyl ethanolamine was cross-correlated against the Raman dataset, further suggesting the particles are indeed localized in a lipid rich environment. Furthermore, to demonstrate the robustness and versatility of the analysis method, a spectrum of pure RNA was used to demonstrate that a differentiation could be made between DNA of the nucleus and RNA of the nucleolus using the supervised spectral cross-correlation technique.