Figure 3

Comparison of mesenchymal-like components between the compound sensitivity and gene expression data sets. Principle component analysis revealed a shared feature in each data set. The second component of the gene expression data, containing 8.4% of the variance, was largely defined by a panel of cell lines identified previously as potentially related to mesenchymal differentiation. The 'mesenchymal' component also appeared to be a significant features of the GI50 data, the cell lines ordered here as in the plot of the gene expression data, though this component explained only 0.9% of the variance. The two 'mesenchymal' components of the gene expression and drug data correlate with each other with an R of 0.62, and therefore represent a strong but readily identifiable correlation. Thus identified, correlations with similarities to these patterns can be noted and subtracted or filtered from analyses.