Supplementary Figure 3: Garnett is robust to downsampling.

A) Boxplots of the percent of cells in 10x Genomic PBMC V1 dataset (n = 94,571 cells) correctly classified at various levels of downsampling of V2 training data. X axis shows the percent of original reads kept after downsampling. At each downsampling level, 10 random downsampling sets were used. (Boxplot elements: center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range; points, outliers). B) Boxplots of the percent of cells in 10x Genomic PBMC V1 dataset (n = 94,571 cells) correctly classified at various levels of downsampling of the V1 test dataset. Classifier was trained on the whole 10x Genomic V2 dataset. X axis shows the percent of original reads kept after downsampling. At each downsampling level, 10 random downsampling sets were used. (Boxplot elements: center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range; points, outliers). C) Boxplots of the percent of cells in 10x Genomic PBMC V1 dataset (n = 94,571 cells) correctly classified when varying the number of T cells included in the V2 training data. X-axis shows the number of T cells included in the training set. At each tested level, 10 random sets of T cells were included. (Boxplot elements: center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range; points, outliers).