Fig. 2
From: The specification game: rethinking the evaluation of drug response prediction for precision oncology
Boxplots showing the distribution of the prediction performances, measured by Pearson’s r values, for the tCNN [16], GraphDRP [15], and the NxtDRP models on the GDSC dataset across three Aggregation Strategies (columns) and three Splitting Strategies (rows A,B, and C). The NxtDRP variants MT, PR, and EX denote the omics data utilized: none, Proteomics, and Transcriptomics, respectively (for further details, refer to Fig. 6B–E)