Figure S3. Benchmark task2: scaling with randomness.

Metric scores by increasingly randomized batch label in a dataset with a moderate batch effect (pbmc_roche). Scores were standardized by subtraction of their mean and division of their SD across permutations. Directions were adjusted when necessary, such that all scores increase with batch strength. Grey lines indicate the scores of the other metrics. Corresponding absolute values of the Spearman correlation coefficients (R) are shown in the text box of each subpanel.