Figure 6

Scores (A) and Correlation loadings (B) of Partial least squares regression using all markers and training samples. Scores (A) and Correlation loadings (B) of the first two components of partial least squares regression (PLSR) of 44 gene markers and the training set of cadmium and phenanthrene treated samples of Folsomia candida. For details about exposure conditions of the samples and qPCR markers, represented by the numbering in 6B (order as in Table 2), see main text. Abbreviations of the samples in 6A: cd = cadmium treatment, ph = phenanthrene treatment, c = control (water and acetone). 6B shows the correlations between the individual x-loadings (gene expressions) and y-loadings (treatments) for the first two components. Circles indicate the completeness of fit of 100% (outer) and 50% (inner) explained variance [35], calculated by cross validation and Jack-knifing. PLSR analysis was done with The Unscrambler multivariate analysis software [67].