“…To benchmark DEGman, we considered eight popular tools, DEsingle [ 7 ], DEseq2 [ 21 ], SigEMD [ 8 ], scDD [ 3 ], edgeR [ 10 ], Monocle2 [ 5 ], glmmTMB [ 12 ] and NEBULA [ 13 ] which have been shown to have superior performance in multiple method comparisons [ 20 , 35 , 61 , 64 , 65 ]. We also compared DEGman with a new DEG finding method, singleCellHaystack [ 14 ], which uses the coordinates of all cells in a low-dimensional space produced by a dimensionality reduction methods, such as principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), or uniform manifold approximation and projection(UMAP) [ 66 , 67 ].…”