Main functionality 2: To select the optimal hyperparameter for metacell partitioning

Usage

mcRigor_OPTIMIZE(
  obj_singlecell,
  cell_membership = NULL,
  assay_type = c("RNA", "ATAC"),
  Gammas = NULL,
  aggregate_method = c("mean", "sum", "geom"),
  output_file = NULL,
  Nrep = 1,
  gene_filter = 0.1,
  feature_use = 2000,
  cor_method = c("pearson", "spearman"),
  prePro = T,
  test_cutoff = 0.01,
  thre_smooth = T,
  thre_bw = 1/6,
  D_bw = 10,
  optim_method = c("tradeoff", "dub_rate_large", "dub_rate_small"),
  weight = 0.5,
  dub_rate = 0.1,
  draw = T,
  pur_metric = NULL,
  check_purity = T,
  fields = NULL,
  step_save = T
)

Arguments

obj_singlecell: Seurat object of the original single cells.
cell_membership: A dataframe, each column of which contains the metacell membership of single cells under a given gamma. The column names should be the corresponding gamma's. The row names should be the single cell id's.
assay_type: The type of data assay yuo are using, depending on which different normalization would be used.
Gammas: The candidate pool of granularity levels to consider in optimization
aggregate_method: The method to aggregate single cell profiles into metacell profiles
output_file: The directory for output saving
Nrep: The number of permutation repetitions we use when deriving the null.
gene_filter: A proportion. Genes expressed lower than this proportion will be filtered out.
feature_use: The number of genes to use in metacell testing.
cor_method: The method for gene correlation calculation description
prePro: A boolean indicating whether to normalize obj_singlecell for preprocessing.
test_cutoff: The test size for dubious metacell detection testing
thre_smooth: A boolean indicating whether to smooth the threshold function
thre_bw: If thre_smooth is True, thre_bw specifies the bandwidth for smoothing.
D_bw: A boolean indicating whether to smooth the dubious rate with respect to metacell size
optim_method: The method used for granularity level optimization. Default is trading off between sparsity and dubious rate
weight: The weight for dubious rate in the tradeoff.
dub_rate: If tradeoff is not used for optimization, what is highest acceptable dubious rate
draw: A boolean indicating whether to visualize the mcRigor results
pur_metric: Can be NULL or a metadata variable name, ex. cell type.
check_purity: A boolean indicating whether to calculate the metacell purity of specific fields or not.
fields: A vector of the fields of interest, ex. celltype. It should be a subset of obj_singlecell's meta.data.
step_save: A boolean indicating whether to save the outputs step by step

Value

A list containing the following fields:

best_granularity_level: The optimal granularity level selected by mcRigor
best_Score: The evaluation score for the metacell partition given by the optimal granularity level selected by mcRigor
opt_metacell: The metacell object build under the optimal granularity level
scores: A data frame containing the evaluation scores for each gamma
optim_plot: The line plot to visualize the tradeoff for hyperparameter opimization.
thre: The thresholds for dubious metacell detection
TabMC: A dataframe containing the permutation results, elements to calculate the test statistics mcDiv and mcDiv null