Fig. 11

Distributions of CLR-transformed HTO UMI counts from a 4-sample cell-hashing experiment. Each distribution is decomposed into two Gaussian distributions, \(\mathcal {N}_{high}(\mu _{high}^{\hat {l}},\,(\sigma _{high}^{\hat {l}})^{2})\) and \(\mathcal {N}_{low}(\mu _{low}^{\hat {l}},\,(\sigma _{low}^{\hat {l}})^{2})\). \(\mathcal {N}_{high}(\mu _{high}^{\hat {l}},\,(\sigma _{high}^{\hat {l}})^{2})\) represents the HTO UMI count distribution of cell-enclosing GEMs. \(\mathcal {N}_{low}(\mu _{low}^{\hat {l}},\,(\sigma _{low}^{\hat {l}})^{2})\) represents the HTO UMI count distribution of cell-free GEMs