simulateData: Simulate RNA-seq bulk gene expression count data

In DavidKLim/FSCseq: Feature Selection and Clustering of RNA-seq Count Data

Description

Simulate data based on input simulation parameters. Size factors are custom input or simulated from N(1,0.25)

Usage

simulateData(
  K = 2,
  B = 1,
  g = 10000,
  n = 50,
  pK = NULL,
  pB = NULL,
  LFCg = 1,
  pDEg = 0.05,
  sigma_g = 0.1,
  LFCb = 1,
  pDEb = 0.5,
  sigma_b = 0,
  beta0 = 12,
  phi0 = 0.35,
  SF = NULL,
  nsims = 25,
  disp = "gene",
  n_pred = 25,
  sim_batch_pred = FALSE,
  LFCb_pred = NULL,
  save_file = TRUE,
  save_dir = NULL,
  save_pref = NULL
)

Arguments

K	integer, number of clusters
B	integer, number of batches
g	integer, number of genes
n	integer, number of samples
pK	vector of length K (optional): proportion of samples in each cluster
pB	vector of length B (optional): proportion of samples in each batch
LFCg	numeric, LFC for cluster-discriminatory genes
pDEg	numeric, proportion of genes that are cluster-discriminatory
sigma_g	numeric, Gaussian noise added to each gene/sample N(0,sigma_g). Default is 0.1
LFCb	numeric, LFC for genes that are differentially expressed across batch. Default is 1.
pDEb	numeric, proportion of genes that are differentially expressed across batch. Default is 0.5.
sigma_b	numeric, batch-specific Gaussian noise (default 0).
beta0	numeric, baseline log2 expression for each gene before LFC is applied
phi0	numeric, baseline overdispersion for each gene
SF	vector of length n (optional), custom size factors from DESeq2. If NULL, simulated from N(1,0.25)
nsims	integer, number of datasets to simulate given the input conditions. Default is 25.
disp	string, either 'gene' or 'cluster' to simulate gene-level or cluster-level dispersions. Default is gene-level. Input phi must be g x K matrix if disp='cluster'
n_pred	integer, number of samples in simulated prediction dataset. Default is 25
sim_batch_pred	boolean: FALSE (no batch effect for prediction samples) or TRUE (batch effect)
LFCb_pred	LFCb for batch-affected genes in prediction set. By default (NULL), = max(batch_effects) + LFCb/2: larger batch effect than training.
save_file	boolean: TRUE (save each set of simulations)
save_dir	string (optional): directory to save files. Default: 'Simulations/<sigma_g>_<sigma_b>/B<B>'
save_pref	string (optional): prefix of file name to save simulated data to. Default: '<K>_<n>_<LFCg>_<pDEg>_<beta0>_<phi0>'

Value

if save_file=TRUE, then saved file in '<save_dir>/<save_pref>_sim<1:nsims>_data.RData'. Otherwise, list of length 'nsims', with a sim.dat list object for each simulation

Author(s)

David K. Lim, [e-mail]

References

https://github.com/DavidKLim/FSCseq

Examples

sim.dat = FSCseq::simulateData(B=1, g=10000, K=2, n=50, LFCg=1, pDEg=0.05, beta0=12, phi0=0.35, nsims=1, save_file=F)[[1]]

DavidKLim/FSCseq documentation built on Dec. 12, 2021, 3:46 a.m.