Development of a Label-free Raman Imaging Technique for Differentiation of Malaria Parasite Infected from Non-Infected Tissue
Abbreviations:
PCA – Principal Component Analysis
PC – Principal component
Figure 1(a): Control (non-infected) and malaria parasite infected tissue were Raman mapped on a Renishaw inVia Raman spectrometer (532 nm). The data had a cosmic ray removal and noise filtering algorithm applied to it in WiRE 4.2 software. Data was then converted to SPC files using WiRE 4.2 software and opened in Matlab 2013a software (The MathWorks, Natick, MA, USA) where it was baselined, truncated, averaged and normalised. The resulting data was then plotted using OriginPro 9.0 (64 bit) software to produce Figure 1(a).
Figure 1(b): The average and normalised control spectra from Figure 1(a) was subtracted from the infected average, normalised spectra (Figure 1(a)) to produce a subtracted Raman spectra. The resulting data was then plotted using OriginPro 9.0 software to produce Figure 1(b).
Figure 1(c):The same control and infected map dataset used for Figure 1(a) was used to create the control peak ratio map (data was un-baselined for the ratio maps). Ratio maps were created using a custom script in Matlab. Dataset was truncated to give the 2 Raman peaks of interest, 748 cm-1 & 1307 cm-1. These were then baselined and the peak intensity subtracted and reshaped to the size of the matrix. The matrix for 748 cm-1 Raman peak was then divided by the matrix for the 1307 cm-1. This gave a ratio data point relating to each spectrum taken in the Raman map. This was plotted to give the control and infected intensity ratio maps shown in Figure 1(c). Both the control and infected maps were a matrix of size 41x41.
Figure 2: In total 8 control (non-infected) and 8 infected Raman maps were taken from different areas and different samples of tissue, of various sizes. This produced a large dataset of Raman spectra for both samples. All data maps had a cosmic ray removal and noise filtering algorithm applied to them in WiRE 4.2 software. They were then converted to SPC files using WiRE 4.2 and opened in Matlab software, where they were baselined and truncated. To have the same number of control and infected spectra to process, 15050 spectra were randomly selected for both using a script in Matlab. These spectra were then averaged in Matlab to give a final 43 control and 43 infected spectra that were used in the empirical analysis. The signal to baseline intensity for the Raman peaks 745 cm-1 & 1307 cm-1 was extracted and the ratio calculated for the 43 control / infected spectra. These were plotted against each other in OriginPro 9.0 (64 bit) to produce Figure 2(a). The average and standard deviation of both data sets were calculated in Matlab and plotted in Origin software to produce Figure 2(b). Unpaired student t-test was performed against the datasets using a script in Matlab.
Figure 3: Dataset used to perform PCA (blue - control; pink - infected). The same 43 control and 43 infected spectra used to perform the empirical analysis (Figure2) were also used in the PCA model. Data originated from 8 control and 8 infected tissue maps, all measured on a Renishaw InVia Raman microscope (532 nm). All maps had a cosmic ray removal and noise filtering algorithm applied in WiRE 4.2 software and were converted to SPC files in WiRE. These were transferred to Matlab software where they were baselined and truncated. 15050 spectra for each tissue type were selected using a random data sampler script in Matlab before being averaged to give 43 final spectra for each tissue type. 43 average control spectra were combined into one matrix with 43 average infected spectra and auto scaled and PCA run using a custom Matlab script. PCA scores plot was plotted in Matlab showing discrimination in the Raman peaks between the control and malaria parasite infected tissue. This is shown in Figure 3(a). Loadings plots for PC1 & PC2 (highlighted in green) were plotted in OriginPro 9.0 software against the Raman shift (cm-1) on the x-axis to produce Figure 3(b).
Figure 4: Solutions of hemoglobin and hemozoin were made and left to dry on a glass slide before being analysed using a Renishaw InVia microscope system. 5 hemoglobin (in blue) and 5 hemozoin (in pink) reference spectra were collected with a 30s acquisition time and 532 nm laser. These were converted to spc files and imported to Matlab R2013a software where they were baselined, averaged and then normalised to the highest peak in each spectrum. The data highlighted in green was plotted in OriginPro 9.0 64 bit software, the spectra were offset when plotted for clarity.
S1: Experimental procedure for the synthesis of β-hematin. Please see supplementary for more information.
S2: Peak intensity ratio maps for a control and infected spleen section. A different infected map dataset to that used for Fig1 was used to create the infected peak ratio map shown in FigS2 but all data was treated the same as in Fig1 (data was un-baselined for the ratio maps). Ratio maps were created using a custom script in Matlab R2013a. Dataset was truncated to give the 2 Raman peaks of interest, 1170 cm-1 & 1307 cm-1. These were then baselined and the peak intensity subtracted and reshaped to the size of the matrix. The matrix for 1170 cm-1 Raman peak was then divided by the matrix for the 1307 cm-1. This gave a ratio data point relating to each spectra taken in the Raman map. This was plotted to give the infected intensity ratio map I1170/I1307 shown in FigS2 [data shown below was plotted].
S3: Empirical analysis of control and infected datasets. The same dataset of 43 control and 43 infected that were used in Fig 1 and Fig 2 were used to create figure S3. The signal to baseline intensity for the Raman peaks 1170 cm-1 & 1585 cm-1 was extracted and the ratio calculated for the 43 infected spectra. These were plotted against control data in OriginPro 9.0 (64 bit) to produce Figure S3(A). The average and standard deviation of the 43 intensity ratios was calculated in Matlab and also plotted in Origin software against the average control to produce Figure S3(B).
S4: Chemical structures of core heme of hemoglobin and hemozoin, drawn using ChemDraw Professional software. Please see supplementary for more information.
S5: Photographs of a control and infected spleen tissue stained using a Prussian blue stain. Please see supplementary for more information.