Figure 3
Untargeted analysis of TMA1 using full scan mass spectral data acquired in the negative ion mode. (a) Unsupervised clustering by tissue types was observed after PCA in the two-dimensional space generated with the first two principal components estimated (68% variance explained). Supervised classification of the samples using all the obtained principal components as features (m = 35) yielded overall 91.7% classification accuracy using a random forest model. (b) The model results are shown in the confusion matrix estimated via six-fold cross-validation. NCB: non-cancerous brain.
