RunBlockCorr

Source code

Description

Run spatial dissimilarity test for features and their binding features in parallel.

Usage

RunBlockCorr(
  object = NULL,
  bind.name = "gene_name",
  features = NULL,
  assay = NULL,
  min.cells = 10,
  bind.assay = NULL,
  bind.features = NULL,
  min.cells.bind = 10,
  prefix = NULL,
  subset = NULL,
  min.features.per.block = 1,
  scale.factor = 10000,
  mode = c(1, 2, 3),
  method = c("D", "D2", "Lee"),
  library.size = NULL,
  wm.name = NULL,
  perm = 100,
  seed = 999,
  threads = 0,
  verbose = TRUE,
  debug = FALSE,
  cells = NULL,
  idents = NULL,
  node = NULL,
  reduction = "pca",
  dims = 1:10,
  k.param = 20,
  prune.SNN = 1/50,
  n.trees = 50,
  nn.eps = 0,
  nn.method = "annoy",
  annoy.metric = "euclidean"
)

Arguments

object Seurat object
bind.name Title name for binding features in the meta table. Consider most users start Yano to perform alternative splicing analysis, the default bind.name set to “gene_name”.
features Vector of features to calculate. Default is AutoCorrFeatures(object).
assay Work assay.
min.cells Features detected in few than minimal number of cells will be skipped. Default is 10.
bind.assay Name of binding assay.
bind.features List of bind features. Default use all.
min.cells.bind Binding features detected in few than minimal number of cells will be skipped. Default is 10.
prefix Prefix name for output scores and values. Default is same with bind.name.
scale.factor Scale factor to normalise counts. Default is 1e4. For mode 1, this function will use data from Layer ‘data’. For mode 2 and 3, will use data from Layer ‘counts’. The counts will further be normalised with sample size and scale factor for spatial dissimilarity test.
mode Test mode. For mode 1, X (test feature) vs Y (binding feature). For mode 2, X vs (Y-X). For mode 3, X vs (Y+X). Please note, when set to mode 2 or 3, will use raw counts to update expression value of binding features. Then normalise the counts before testing. For mode 1, will use Layer ‘data’. Default is mode 1.
method Method to use. Support D, D2 and Lee. D = sqrt(Lx)(1 − rxy). D2 = sqrt(Lx)*sqr**t(Ly)(1 − rxy*). Lee for Lee’s Score. In default use D method, see the manual for details.
library.size Library size for each cell, used for normalise counts when mode is 2 or 3. If not set, use colSum(counts) instead.
wm.name Weight matrix name, this matrix (graph) generated by RunAutoCorr.
perm Permutations for evaluating mean and sd of D/L scores. Default is 100.
seed Seed for generate random number. Default is 999.
threads Threads. If set to 0 (default), will auto check the CPU cores and set threads = number of CPU cores -1.
debug Print debug message. Will auto set thread to 1. Default is FALSE.
cells Calculate scores for predefined cells. Will reconstruct the SNN graph and weight matrix for these cells with ‘reduction’ space (usually be pca or harmony). Only weight matrix that calculated by SNN is supported if cells/idents/node is defined.
idents Calculate scores for these cell groups. The idents should be a vector of group names in Idnets(object).
node A node to find markers for and all its children; requires BuildClusterTree to have been run previously. Only can be used if test all groups.
reduction Dimension reduction name for constructing SNN graph and weight matrix. Default is ‘pca’. This and following parameters only actived when cells is set, because need to recalculate the SNN graph for the defined cells.
dims Dimensions of reduction used to construct SNN graph.
k.param Defines k for the k-nearest neighbor algorithm.
prune.SNN Sets the cutoff for acceptable Jaccard index when computing the neighborhood overlap for the SNN construction. Any edges with values less than or equal to this will be set to 0 and removed from the SNN graph. Essentially sets the stringency of pruning (0 — no pruning, 1 — prune everything). Default is 1/50.
n.trees More trees gives higher precision when using annoy approximate nearest neighbor search. Default is 50.
nn.eps Error bound when performing nearest neighbor seach using RANN; default of 0.0 implies exact nearest neighbor search
nn.method Method for nearest neighbor finding. Options include: rann, annoy(default).
annoy.metric Distance metric for annoy. Options include: euclidean (default), cosine, manhattan, and hamming
versbose Print log message. Default is TRUE.
Back to top