Figure 1

Assessment of systematic variability across plates, chips, and spots using principal component analysis (PCA) on the 144 quality control samples used in the liver cancer and liver toxicity study. The 144 spectra obtained from assays of quality control samples from the liver cancer and liver toxicity study were first normalized. The five peaks which appeared in all 144 spectra were taken as the common proteins and used for PCA analysis using software Spotfire DecisionSite 7.1 (Somerville, MA, http://www.spotfire.com). (A) The PCA results were color coded for the six Ciphergen bioprocessor plates used to assay the samples. The points represent the spectra and are spread randomly throughout the space. The spectra from each plate did not group together indicating a lack of systematic variability. (B) The PCA results were color coded for spot position. Each protein chip has eight spots. The spectra obtained from samples loaded on any same spot position can not be grouped together and randomly spread around in the PCA figure. Thus, spot position did not introduce systematic variability in the data. (C) The PCA results were color coded for the chip number in which each sample was loaded for SELDI analysis. Total 12 chips were used for one bioprocessor plate. The spectra of samples from any chips were not grouped together in the PCA, but rather were spread randomly in the space, indicating a lack of systematic variability due to the chip used to assay samples.