Two‐dimensional correlation analysis of Raman microspectroscopy of subcellular interactions of drugs in vitro

Abstract

Two‐dimensional (2D) correlation analysis is explored to data mine the time evolution of the characteristic Raman microspectroscopic signatures of the subcellular responses of the nucleoli of human lung cancer cells to the uptake of doxorubicin. A simulated dataset of experimental control spectra, perturbed with systematically time‐dependent spectral changes, constituted by a short‐term response which represents the initial binding of the drug in the nucleolus, followed by a longer term response of the organelle metabolism, is used to validate the analysis protocol. Applying 2D correlation analysis, the in phase, synchronous correlation coefficients are seen to contain contributions of both response profiles, whereas they can be independently extracted from the out of phase, asynchronous correlation coefficients. The methodology is applied to experimental data of the uptake of doxorubicin in human lung cell lines to differentiate the signatures of chemical binding and subsequent cellular response.

Data mining Raman microspectroscopic responses of cells to drugs in vitro using multivariate curve resolution-alternating least squares

“…Here, we provide a summary of the materials and methods used for the experimental data. Further information can be accessed in the original publications [17,20]. to a total data set of 1260 spectra.…”

“…Experimental control spectra were derived from the study of Farhane et al [20] , measured as described above, and the 30 nucleolar spectra averaged. Based on the results of the PLSR analysis of the original study, simulated early (Response 1) and late (Response 2) spectral response profiles were generated in the Horiba Labspec 5 software, as described in Byrne et al [17] The temporal evolution of the respective contributions of the simulated responses were derived according to the rate equations described in the "Results Section". Simulated datasets were created by adding weighted sums of the responses to the control spectra using the Matlab 2019a platform (Mathworks, Natick, MA, USA).…”

“…Simulations were based on previous works by Farhane et al [20] and Byrne et al [17], which identified spectral The effects of drug uptake after inoculation on a cell, described by the intensity of Responses 1 and 2, can be simulated considering ordinary differential equations [17]. , the number of DOX molecules bound to DNA and , the intensity of the subsequent cellular response as a function of time can be expressed as:…”

“…In summary, the results indicate that MCR-ALS outperforms the OPLSR method in extracting kinetic information as well as spectral markers of the different processes involved. On the other hand, in a recent study using the same experimental data and simulation approach, the use of a two dimensional correlation analysis [17] method was also able to determine that bands associated with Response 1 and Response 2 are asynchronous (i.e. they occurred at different rates).…”

“…[16] Although the performance of OPLSR models is generally largely equivalent to that provided by PLSR, OPLSR models gain in terms of interpretability of the variables, all the variance associated with Y being represented in the first latent variable (LV). Other methods to investigate these systems include two-dimensional [17] correlation analysis and Principal Component Analysis [18].…”