gemini_initialize: gemini_initialize

In gemini: GEMINI: Variational inference approach to infer genetic interactions from pairwise CRISPR screens

Description

Creates a gemini.model object given Input data and initialization parameters.

Usage

gemini_initialize(Input, guide.pair.annot = "guide.pair.annot",
  replicate.map = "replicate.map", sample.column.name = "samplename",
  LFC.name = "LFC", nc_gene, CONSTANT = 32, concordance = 1,
  prior_shape = 0.5, pattern_join = ";", pattern_split = ";",
  window = NULL, monotonize = FALSE, verbose = FALSE, cores = 1,
  save = NULL)

Arguments

Input	an object of class gemini.input
guide.pair.annot	the name of an object within Input containing guide to gene annotations (Default = "guide.pair.annot")
replicate.map	the name of an object within Input containing replicate and sample mappings (Default = "replicate.map")
sample.column.name	a character or integer indicating which column of Input$replicate.map describes the samples.
LFC.name	a character indicating an object within Input to treat as LFC. By default, "LFC" is used, which is the output of gemini_calculate_lfc.
nc_gene	a character naming the gene to use as a negative control. See details for more.
CONSTANT	a numeric indicating a constant value that shifts counts to reduce outliers, see Details (default = 32).
concordance	a numeric value to initialize x (default = 1)
prior_shape	shape parameter of Gamma distribution used to model the variation in the data in Input (default = 0.5)
pattern_join	a character to join the gene combinations found in guide.pair.annot. Default ';'
pattern_split	a character to split the guide combinations found in guide.pair.annot. Default ';'
window	a numeric if window smoothing should be done on initialized tau values, otherwise NULL (default) for no window smoothing
monotonize	a logical specifying whether the variance should be monotonically increasing. (default FALSE)
verbose	default FALSE
cores	numeric indicating the number of cores to use. See details in gemini_parallelization.
save	a character (file path) or logical indicating whether the initialized model should be saved.

Details

guide.pair.annot is created with the following format:

• 1st column contains guide pairs, joined by pattern_split.

• 2nd column contains the name of gene mapping to the first guide in the 1st column.

• 3rd column contains the name of gene mapping to the second guide in the 1st column.

Additional columns may be appended to guide.pair.annot, but the first three columns must be defined as above.

Use of negative control As described in Zamanighomi et al., it is highly recommended to specify a negative control. In the event that no negative control is specified, GEMINI will use the median LFC of all guide pairs targeting a gene to initialize that gene's effect. While this may be reasonable in large all-by-all screens, it is not recommended in smaller screens or some-by-some screens. As a result, when possible, be sure to specify a negative control.

Value

a Model object of class gemini.model

Examples

data("Input", package = "gemini")
Model <- gemini_initialize(Input, nc_gene = "CD81")

gemini documentation built on Nov. 8, 2020, 8:22 p.m.