Figure 5
Leave-one-out Cross Validation of the elastic net regression model based on the primary (top) and secondary (bottom) screens for normal and cancer cell lines after logarithmic transformation on the data. Each of the 140 point in these figures corresponds to one of the 140 drugs. “Regression” refers to –log of the viability predicted by the regression model using all data from the other 139 drugs as training set, while “Measured” refers to –log of the actual viability measured for the drug or drug combination. Note that, as in Figure 4, only the secondary screen leads to predictive models with significant R 2 for both cell types. The R 2 for the Cancer cell lines is considerably better using the log transformation.