Fig 3. Augmenting with scores at different assay conditions did not improve predictions.

At 15% mutational completeness, the data size was augmented by combining data from six different assay conditions. There was no improvement in the prediction quality although the data was enhanced 6-fold. The $R_{test}^2$ with and without data augmentation was 0.61 and 0.66 respectively. More detailed results are in Fig 10 in S2 File. Similar analysis was performed by developing model trained on data from the 650 μg/ml drug concentration assay. In this case also the predictive ability of the models trained at only one concentration or at multiple concentrations was similar.