Running intra-metacell clustering on each metacell, identifying splits if existing. Also remove cells with outlier expression from each metacell. Note that this can in many case eliminate doublets as outliers - but may be insffucieint as doublets may constitute homogeneous metacells in large datasets.

mcell_mc_split_filt(
new_mc_id,
mc_id,
mat_id,
T_lfc,
plot_mats = T,
dirichlet = F
)

## Arguments

new_mc_id

id of metacells to create

mc_id

id of source metacell object

mat_id

id of umi matrix

plot_mats

if this is true, a heatmap will be generated for each metacell

dirichlet

if this is true, the distance metric to be used for splitting is dirichlet on the downsampled umi counts. Otherwise (and by default), the correlation between log(1+umi) of the downsampled matrix will be used. Note that downsampling is done given the miminum depth of each metacell, seperately.