simulateRandomWalk: Simulate random walks

In farrellja/URD: URD

Description

In order to find each developmental trajectory, biased random walks are used to find the paths through the data that connect particular tips to the root. This function simulates random walks from a given starting population (usually cells in a tip) to a given ending population (usually the cells in the root), using connections in the provided transition matrix (usually biased by pseudotimeWeightTransitionMatrix). The output is then turned into visitation frequency by the function processRandomWalks.

Usage

simulateRandomWalk(
  start.cells,
  transition.matrix,
  end.cells,
  n = 10000,
  end.visits = 1,
  verbose.freq = 0,
  max.steps = 5000
)

Arguments

start.cells	(Character vector) Cells to use as a starting pool. One cell is chosen at random each simulation.
transition.matrix	(Matrix or Sparse Matrix) Transition matrix (generally biased by pseudotimeWeightTransitionMatrix.)
end.cells	(Character vector) Stop the random walks when they visit these cells. (Usually the root cells.)
n	(Numeric) Number of walks to simulate
end.visits	(Numeric) Number of visits to end.cells to do before stopping
verbose.freq	(Numeric) Print a progress update, every verbose.freq walks. If 0, no progress updates are reported.
max.steps	(Numeric) Maximum number of steps to take before aborting the walk, making the assumption that the walk has somehow gotten stuck. Returns NULL for those walks.

Details

This function can be accelerated about 10-fold by using a full matrix of transition probabilities, rather than a sparse one, though at the cost of using large amounts of RAM for data sets with many cells.

Value

(Character vector) Names of cells visited during the random walk.

Examples

# Determine parameters of logistic function
diffusion.logistic <- pseudotimeDetermineLogistic(object, "pseudotime", optimal.cells.forward = 40, max.cells.back = 80, pseudotime.direction = "<", do.plot = T, print.values = T)

# Generate biased transition matrix
biased.tm <- pseudotimeWeightTransitionMatrix(object, pseudotime = "pseudotime", logistic.params = diffusion.logistic, pseudotime.direction = "<")

# Simulate random walks
these.walks <- simulateRandomWalk(start.cells = tip.10.cells, transition.matrix = biased.tm, end.cells = root.cells, n = 50000, end.visits = 1, verbose.freq = 2500, max.steps = 5000)

# Process walks into visitation frequency
object <- processRandomWalks(object, walks = these.walks, walks.name = "10", verbose = F)

farrellja/URD documentation built on June 17, 2020, 4:48 a.m.