runICA: Custom ICA function for analyzing gene expression data.

Description Usage Arguments Value Examples

View source: R/runICA.R

Description

Performing ICA on a dataset and create a list object with results.

Usage

1
2
3
runICA(pheno_mx = NULL, assay_idx = NULL, k_est = NULL, var_cutoff = 99,
  n_runs = 1, n_cores = NULL, scale_pheno = FALSE, h_clust_cutoff = 0.3,
  max_iter = 10, random_seed = NULL, similarity_measure = "peaks")

Arguments

pheno_mx

Phenotype matrix with diemnsions g x N, where g is the number of genes and N is the number of samples.

assay_idx

The assay index to be used to retrieved a single assay from the SummarizedExperiment object.

k_est

Number of components to be estimated or method to estimate it.

var_cutoff

Percent variance threshold to use when <k_est> is not supplied.

n_runs

Number of times to run ICA. If this is set to a number larger than 1, only ICs that replicate between runs are going to be returned

n_cores

Number of cores to use when n_runs is larger than 1.

scale_pheno

If set to TRUE the pre-processing step of the data will include a scaling step. This will divide all phenotypes by their standard deviation.

h_clust_cutoff

is the cutoff value used in hierarchical clustering. Default is set to 0.3.

max_iter

Maximum iterations for estimating k for each run. Default value is set to 10.

random_seed

Set a specific value for random seed to reproduce the results of ICA.

similarity_measure

How to measure the similarity between ICs. If set to "peaks" only gene weights that are greater than 1 sd are used to calculate similarity.

Value

The following entries will be generated in the output list ica_object after running the example above.
A : The IC coefficient matrix, with each row representing coefficients for the corresponding independent component. (standard fastICA() output)
S : Matrix of gene weights for each independent component. Each column holds a single component. (standard fastICA() output)
percent_var : The percent variance each independent component is explaining.
peaks : Indicating which gene has a gene weight larger than 2 standard deviations of its component gene weights.
order : The order of independent components based on the variance that they explain.
X, K, W : Standard outputs of fastICA(). X is the pre-processed data matrix, K is the pre-whitening matrix projecting the data onto the first n principal components, and 'W' is the estimated unmixing matrix.
Three attributes are set within the list object. "ICAobject" for class, "ica" for method and "no" for covar_cor.

Examples

1
2
3
data(expr_data)

ica_result <- runICA(expr_data)

jinhyunju/picaplot documentation built on May 19, 2019, 10:35 a.m.