Fig. 4
Biological interpretability in outliers between actual and predicted protein abundances. a Scatterplots of measured protein abundance (y) against L1-regularized predicted (\(\hat y\)), using gradient-boosted regression trees (GBRT) with LOOCV for G1, S and G2/M cell cycle. 90th percentile outliers with respect to the squared-error (ε) highlighted in red. no refers to the number of outliers. b-c Venn diagrams of outlier (red) overlap between RFE, L1 and K-Best feature selectors per cell cycle phase (b), and between G1, S and G2/M cell cycle phases per feature selector (c)