Score plot of first principal component (PC1) versus second principal component (PC2) from partial least square (PLS) regression analysis of tumor cell density to spin-echo spectral data (A), the corresponding loading profile of PC1 (B), and the predicted versus measured tumor cell density (C). In (A) and (C) each point represents the data of a single biopsy. In (A) the color code for tumor cell densities (cells/mm2) is shown. Spectral assignments are abbreviated in (B): β-Glc, β-glucose; Lac, lactate; m-Ino, myo-inositol; Cre, creatine; Tau, taurine; GPC, glycerophosphocholine; PC, phosphocholine; Cho, choline. The Pearson correlation coefficient and p-value are marked in (C). Total residual y-variance and root mean square error of prediction were minimised by retaining 3 PCs in the model. These three PCs accounted for 52% of the total x-variation, and 65% of the total y-variation.