Quantile align (normalize) factor loadings

This is a deprecated function. Calling 'quantileNorm' instead.

Usage

quantileAlignSNF(
  object,
  knn_k = 20,
  k2 = 500,
  prune.thresh = 0.2,
  ref_dataset = NULL,
  min_cells = 20,
  quantiles = 50,
  nstart = 10,
  resolution = 1,
  dims.use = 1:ncol(x = object@H[[1]]),
  dist.use = "CR",
  center = FALSE,
  small.clust.thresh = 0,
  id.number = NULL,
  print.mod = FALSE,
  print.align.summary = FALSE
)

Arguments

object

liger object. Should run optimizeALS before calling.

knn_k

Number of nearest neighbors for within-dataset knn graph (default 20).

k2

Horizon parameter for shared nearest factor graph. Distances to all but the k2 nearest neighbors are set to 0 (cuts down on memory usage for very large graphs). (default 500)

prune.thresh

Minimum allowed edge weight. Any edges below this are removed (given weight 0) (default 0.2)

ref_dataset

Name of dataset to use as a "reference" for normalization. By default, the dataset with the largest number of cells is used.

min_cells

Minimum number of cells to consider a cluster shared across datasets (default 2)

quantiles

Number of quantiles to use for quantile normalization (default 50).

nstart

Number of times to perform Louvain community detection with different random starts (default 10).

resolution

Controls the number of communities detected. Higher resolution -> more communities. (default 1)

dims.use

Indices of factors to use for shared nearest factor determination (default 1:ncol(H[[1]])).

dist.use

Distance metric to use in calculating nearest neighbors (default "CR").

center

Centers the data when scaling factors (useful for less sparse modalities like methylation data). (default FALSE)

small.clust.thresh

Extracts small clusters loading highly on single factor with fewer cells than this before regular alignment (default 0 -- no small cluster extraction).

id.number

Number to use for identifying edge file (when running in parallel) (generates random value by default).

print.mod

Print modularity output from clustering algorithm (default FALSE).

print.align.summary

Print summary of clusters which did not align normally (default FALSE).

Value

liger object with H.norm and cluster slots set.

Details

This process builds a shared factor neighborhood graph to jointly cluster cells, then quantile normalizes corresponding clusters.

The first step, building the shared factor neighborhood graph, is performed in SNF(), and produces a graph representation where edge weights between cells (across all datasets) correspond to their similarity in the shared factor neighborhood space. An important parameter here is knn_k, the number of neighbors used to build the shared factor space (see SNF()). Afterwards, modularity-based community detection is performed on this graph (Louvain clustering) in order to identify shared clusters across datasets. The method was first developed by Waltman and van Eck (2013) and source code is available at http://www.ludowaltman.nl/slm/. The most important parameter here is resolution, which corresponds to the number of communities detected.

Next we perform quantile alignment for each dataset, factor, and cluster (by stretching/compressing datasets' quantiles to better match those of the reference dataset). These aligned factor loadings are combined into a single matrix and returned as H.norm.

Examples

if (FALSE) {
# liger object, factorization complete
ligerex
# do basic quantile alignment
ligerex <- quantileAlignSNF(ligerex)
# higher resolution for more clusters (note that SNF is conserved)
ligerex <- quantileAlignSNF(ligerex, resolution = 1.2)
# change knn_k for more fine-grained local clustering
ligerex <- quantileAlignSNF(ligerex, knn_k = 15, resolution = 1.2)
}