Fig. 4
From: Picky with peakpicking: assessing chromatographic peak quality with simple metrics in metabolomics
False discovery rate and proportion of total good features identified as good by the two-parameter model trained on the combined MESOSCOPE/Falkor dataset and applied to each dataset. A stricter likelihood threshold is used here (0.9) than in Fig. 2. FDR is calculated by dividing the number of false positives by the total positives produced by the model and GFF is calculated by dividing the number of true positives by the total number of good features as identified manually (only possible for fully-labeled datasets). Points correspond to the calculated percentage and absolute numbers are provided above/below the point