inferNetwork: Infers a Gene Regulatory Network from Time Series Data

In tianyu-lu/dynUGENE: Dynamical Uncertainty-aware GEne NEtwork inference with random forests

Description

Given a dataframe of p genes as columns and measurements at different timestamps as rows, returns the adjacency matrix of the inferred network, the estimated decay rates of each species, and the dynamics of the network learned by p random forests.

Usage

inferNetwork(
  data,
  multipleExp = FALSE,
  mask = NULL,
  ntree = 10L,
  mtry = NULL,
  alpha = "fromData",
  seed = 777,
  showPlot = FALSE,
  showScores = TRUE
)

Arguments

data	A data.frame of gene expression values, should be numerics. Each row is a measurement of all genes in the system at the indicated time point given in the first column entry of that row. Time stamps do not need to be regularly spaced. Subsequent columns are the gene concentrations measured at the corresponding time stamps. If multiple time series are included, they must be concatenated as new rows, where the first time stamp for the new experiment is less than the last time stamp of the previous experiment.
multipleExp	Optional. Defaults to FALSE. When TRUE, data will be taken to have multiple experiments.
mask	A matrix which only includes the values 1 or NA. Must be of size numgenes*numgenes. If entry (i.j) = 1, then i can be used in predicting the value of j. Otherwise, the connection is snipped and such a dependency is not allowed when training the random forests.
ntree	A positive integer indicating the number of trees in each random forest. Equivalent to the ntree argument in the randomForest package. Defaults to 10L.
mtry	A positive integer indicating the number of randomly sampled candidates to use at each split of each random forest. Equivalent to the mtry argument in the randomForest package. Defaults to p/3, where p is the number of genes. This option is disabled when a mask is provided and the default value is used.
alpha	Identical to the alpha argument in dynGENIE3: Can be "fromData" (default), or a vector containing the gene degradation rates, or a single number. When alpha is "fromData", the degradation rate of each gene is estimated from the data, by assuming an exponential decay between the highest and lowest observed expression values. When alpha is a single number, all the genes are assumed to have the same degradation rate alpha.
seed	Random seed for reproducibility. Defaults to 777.
showPlot	Plots the weights matrix as a heatmap. Defaults to FALSE.
showScores	Show the importance scores when showPlot is set to TRUE. Defaults to TRUE.

Value

Returns an object of class "ugene" with the following items:

• network - A matrix storing the importance weights w_ij of each pair of genes.

• alpha - A vector of the gene product degradation rates, possibly inferred from data.

• model - A list of "randomForest" objects where model[i] is the trained randomForest able to predict changes in concentrations of gene i given the current concentrations of all genes.

References

Geurts, P. (2018). dynGENIE3: dynamical GENIE3 for the inference of gene networks from time series expression data. Scientific reports, 8(1), 1-12.

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18–22.

Examples

## Not run: 
# Infer network from provided repressilator data
ugene <- inferNetwork(Repressilator, showPlot = TRUE)

# Stochastic repressilator data
ugene <- inferNetwork(StochasticRepressilator, multipleExp = TRUE)

## End(Not run)

tianyu-lu/dynUGENE documentation built on Jan. 7, 2021, 6:27 p.m.