Fig. 7From: Benchmarking differential abundance methods for finding condition-specific prototypical cells in multi-sample single-cell datasetsHyperparameter sensitivity of the six DA testing methods. Performance of the four DA testing methods (DA-seq, Meld, Cydar, and Louvain) on the three synthetic datasets and the COVID-19 scRNA-seq dataset with a range of DA ratios (0.75, 0.85, and 0.95) in the target DA cell population. The boxplots show (e.g., each data point) the distribution of AUROC scores across various hyperparameters. High variance implies sensitivity to choice of hyperparameters. #points represents the number of points for each boxplotBack to article page