scandal_programs_of_intra_sample_heterogeneity: Programs of intra-sample heterogeneity
In dravishays/scandal: A framework for single-cell analysis

Description Usage Arguments Details Author(s)

This function extracts transpriptomic programs

scandal_programs_of_intra_sample_heterogeneity(
  object,
  samples = NULL,
  clustering_data = NULL,
  algorithm = "nmf",
  rank = 10,
  ngenes1 = 50,
  ngenes2 = 30,
  sd_threshold = 0.8,
  filter_method = "relative",
  n_clusters_min = 2,
  n_clusters_max = 10,
  bin_control = TRUE,
  n_control_bins = 25,
  n_bin_genes = 100,
  return_all = FALSE,
  verbose = FALSE,
  ...
)

...

further arguments passed to nmf function

Generally, the algorithm extracts programs in a bottoms-up approach starting from extracting programs within each individual sample. The algorithm then detecs highly variable programs that best represent coherent program clusters and aggregates these clusters into programs that generalize accross all samples.
The algorithm performs the following steps:

Call prepare_samples to prepare a ScandalDataSet for each individual sample. This step can be bypassed by supplying a valid samples argument.
For each individual sample generate a predefined number of clusters (configurable by the rank parameter). At the moment the only supported clustering algorithm is NMF via nmf_run. This step can be bypassed by supplying a valid clustering_data argument.
For each sample extract an initial list of within-sample programs by calling nmf_extract_programs. The number of genes in each program is configurable by the ngenes1 parameter. The number of initial within-sample programs correpsonds to rank.
For each sample score the cells for each of the within-sample programs by calling score_within_samples. See scalop::score for more details about how to score cells while controling for differences in cell complexities.
For each sample compute the standard deviation of scores for each within-sample program by calling compute_programs_sd.