Table 1.
Cytokines Analysed.
Table 2.
Number of Principal Components Derived and Proportion of Variation Explained.
Figure 1.
48 Hour Time Pools Principal Component Analysis.
The figure shows cerebral microdialysis derived cytokine data from 12 patients (A–L) pooled into 48 hour time epochs(1–4). Principal component analysis has been used to identify the first 2 principal components which explain 63.7% of the variation in the dataset. Part A is a scores plot which shows the scores on each principal component for each of the observations. Part B is a loading plot which shows the cytokines which load on the respective principal components. Functionally related cytokines, such as IL1b and TNF, cluster within the same quadrant of the plot suggesting that they co-vary.
Figure 2.
48 Hour Time Pools Partial Least Squares Discriminant Analysis.
The figure shows the same cerebral microdialysis derived cytokine data as in figure 1, from 12 patients (A–L) pooled into 48 hour time epochs (Red 0–48 hours, Green 48–96 hours, Blue 96–144 hours, Yellow 144–192 hours). Partial Least Squares Discriminant Analysis is a regression extension of Principal Component Analysis in which the model identifies the greatest sources of variation between pre-specified groups of observations. In this case the supervising variable is time. There is a clear shift in the pattern of observations in the scores plot (part A) over time from the 0–48 hour epoch (red) to the 48–96 hour time epoch (green) to the later time points (blue and yellow). The loading plot (part B) illustrates the cytokines that are responsible for the pattern apparent in part a.
Figure 3.
72 Hour Time Pools Principal Component Analysis.
The figure shows cerebral microdialysis derived cytokine data from 12 patients (A–L) pooled into 72 hour time epochs(1,2). Principal component analysis has been used to identify the first 2 principal components which explain 55.6% of the variation in the dataset. Part A is a scores plot which shows the scores on each principal component for each of the observations. Part B is a loading plot which shows the cytokines which load on the respective principal components. In a similar pattern to figure 1, IL1b and TNF still cluster within the same quadrant of the plot.
Figure 4.
72 Hour Time Pools Partial Least Squares Discriminant Analysis.
The figure shows the same cerebral microdialysis derived cytokine data as in figure 3, from 12 patients (A–L) pooled into 72 hour time epochs (Red 0–72 hours, Green 72–144 hours). Partial Least Squares Discriminant Analysis is a regression extension of Principal Component Analysis in which the model identifies the greatest sources of variation between pre-specified groups of observations. In this case the supervising variable is time. There is a clear shift in the pattern of observations in the scores plot (part A) over time from the early time points (red) to the later time points (green). The loading plot (part B) illustrates the cytokines that are responsible for the pattern apparent in part a.
Figure 5.
Paired Microdialysis and Plasma Samples Principal Component Analysis.
The figure shows paired microdialysate and plasma derived cytokine data from 12 patients. Principal component analysis has been used to identify the first 2 principal components which explain 33.4% of the variation within the dataset. Part A is a scores plot which shows the scores for each observation on each of the principal components and even in this unsupervised model a clear separation between microdialysis and plasma derived cytokines is apparent. Part B is a loading plot which shows the cytokines responsible for the differences between the two biological compartments.
Figure 6.
Paired Microdialysis and Plasma Samples Partial Least Squares Discriminant Analysis.
This figure shows paired microdialysate and plasma derived cytokine data from 12 patients as in figure 5. In this instance the source of the sample (microdialysate vs plasma) has been used as a supervising variable within a partial least squares discriminant analysis to identify two principal components that maximise the differentiation between the two biological compartments. Part A is a scores plot which shows the scores on each principal component for each observation. As with figure 5, there is a clear separation between the two biological compartments. Part B is a loading plot which demonstrates which cytokines are responsible for the difference apparent in part a. The cytokines on the left side of the figure show a comparative neural tropism when the two biological compartments are compared.