Skip to main content

Table 2 scRATE classification of genes in the heart data

From: Bayesian model selection reveals biological origins of zero inflation in single-cell transcriptomics

Threshold Selected model
  P NB ZIP ZINB
(a)
0 SE 1111 2930 525 949
1 SE 2112 3183 81 139
2 SE 2930 2509 5 71
3 SE 3445 2035 1 34
(b)
0 SE 1523 2317 518 1105
1 SE 2733 2583 63 84
2 SE 3544 1913 3 3
3 SE 4000 1461 0 2
(c)
0 SE 1118 2745 478 1094
1 SE 2107 3108 88 132
2 SE 2930 2432 5 68
3 SE 3420 1981 1 33
  1. Genes were classified as one of four count models (P, NB, ZIP, or ZINB) using four levels of stringency (0 SE, 1 SE, 2 SE, or 3 SE). Table shows the number of genes in each category using a GLM with only the offset term to account for cell sequencing depth (a), using a GLM that also includes cell type as an explanatory covariate (b), and using a GLM that includes offset as well as a randomly shuffled cell type as a covariate (c). See Additional file 1: Tables S3 and S4 for the results with the mouse kidney and the human PBMC datasets