MultiPower: Optimal sample size estimation and power study.

In ConesaLab/MultiPower: Estimation of sample size in multi-omic experiments

Optimal sample size estimation and power study.

Description

MultiPower computes the optimal sample size for a multi-omic experiment when pilot multi-omic data sets are available for estimating the parameters required to compute power. An optimization problem is solved in order to achieve the desired power while minimizing the cost of the experiment.

Usage

MultiPower(data, groups, type, omicPower = 0.6, averagePower = 0.85, null.effect = 0,
fdr = 0.05, cost = 1, equalSize = TRUE, max.size = 200, omicCol = NULL, powerPlots = TRUE)

Arguments

data	List with as many elements as omic data types. The names of the omics should be the names of the list. Each element in this list must be a raw count data matrix, and in this case MultiPower will take into account the library sizes to estimate power; a normally distributed data matrix which must have been already pre-processed and normalized; or a binary data matrix (with 0/1 or TRUE/FALSE values). In any case, for each one of these matrices, rows must correspond to omic features (genes, methylation sites, ChIP-seq peaks, etc.) and columns to observations (biological samples, patients, etc.).
groups	List with as many elements as omic data types. The names of the omics should be the names of the list. Each element in this list must be a vector with length equal to the number of observations for that omic in data argument. Each element of this vector must indicate the experimental group where each observation belong. Only two experimental groups are allowed.
type	Vector with length equal to the number of omic data types. Each element of this vector must be a 1, 2 or 3 to indicate whether the omic data are count data (1), continuous data approximately following a normal distribution (2) or binary data (3).
null.effect	Value of the effect size that corresponds to null hypothesis. By default, 0.
omicPower	The minimum power that must be achieved for each omic. It must be a vector with length equal to the number of omics. If it is a single number, this same number will be used for all the omics. By default, omicPower = 0.6.
averagePower	The minimum average power that must be globally achieved. By default, averagePower = 0.85.
fdr	False Discovery Rate level to be used. It is the significance level after multiple testing correction. By default, fdr = 0.05. If no multiple testing correction is to be applied, this argument must be set to NULL and then alpha argument is required.
cost	The cost to generate a replicate (a sample) for each omic. It must be a vector with length equal to the number of omics. If it is a single number, this same number will be used for all the omics. This argument will only be used when a different sample size per omic is allowed. By default, cost = 1 (which means that all the omics will be assumed to have the same cost).
equalSize	If TRUE (default), the same optimal sample size will be estimated for all the omics. If FALSE, omics are allowed to have different sample sizes.
max.size	Maximum allowed sample size. By default, max.size = 200.
omicCol	The color that will be used to plot each omic. It must be a vector with length equal to the number of omics. If it is NULL (default), default colors are used.
powerPlots	If TRUE (default), power plots will be generated.

Value

When applying MultiPower, the result is a list containing the following elements:

parameters	List with as many elements as omic data types. For each omic, each element of the list is another list containing the different parameters used to compute power, either estimated from the pilot data or provided by the user: type, pooledSD, d, delta, logFC, mu, m, etc.
optimalSampleSize	List containing the following elements: n0 (sample size to achieve the minimum omic power, omicPower, for each omic), n (optimal sample size), finalPower (power at the optimal sample size for each omic), fdr (see fdr argument), omicPower (see omicPower argument), averagePower (see averagePower argument), and cost (see cost argument).
summary	Table summarizing MultiPower results. The columns are: the names of the omic data sets (omic), the omic data type (type), the number of omic features for each omic (numFeat), the minimum and maximum observed Cohen’s d (minCohenD and maxCohenD), the FDR value (FDR), the minimum power to be achieve for each omic (minPower), the average power to be achieved in the multi-omic experiment (averPower), the cost per omic (cost), the minimum sample size needed for each omic to achieve minPower (minSampleSize), the optimal sample size (optSampleSize), and the power at this optimal sample size (power).
data2plot	Data generated to create the power plots that are also returned by the function.

Author(s)

Sonia Tarazona; David Gómez-Cabrero

ConesaLab/MultiPower documentation built on April 16, 2023, 11:39 a.m.