Figure 3
Robustness benchmark for (A) different sample sizes and (B) noise levels. First set of box plots (purple) shows the overall performance,\({R}^{2}\)(log scaled y axis) of inferred gene models using all available information layers. All following sets describe the stand-alone performances of Proteome, Methylation, Mutation, Clinical and Transcriptomic information layers. 1st order links (green) and 2nd order links (orange) are benchmarked separately. Note, Clinical and Transcriptome information consists of only 1st and 2nd order links. (A) Data sets with different sample sizes were generated using 5–100% of the 604 breast invasive carcinoma samples. (B) Different test data sets were simulated by adding Gaussian noise with increasing variance to the scaled feature values. Here, the noise level reflects the \({R}^{2}\) for six intensities.
