lm_adjust: Linear Adjustment Normalization

In YosefLab/scone: Single Cell Overview of Normalized Expression data

Linear Adjustment Normalization

Description

Given a matrix with log expression values and a design matrix, this function fits a linear model and removes the effects of the batch factor as well as of the linear variables encoded in W.

Usage

lm_adjust(log_expr, design_mat, batch = NULL, weights = NULL)

Arguments

log_expr	matrix. The log gene expression (genes in row, samples in columns).
design_mat	matrix. The design matrix (usually the result of make_design).
batch	factor. A factor with the batch information, identifying batch effect to be removed.
weights	matrix. A matrix of weights.

Details

The function assumes that the columns of the design matrix corresponding to the variable for which expression needs to be adjusted, start with either the word "batch" or the letter "W" (case sensitive). Any other covariate (including the intercept) is kept.

Value

The corrected log gene expression.

Examples

set.seed(141)
bio = as.factor(rep(c(1,2),each = 2))
batch = as.factor(rep(c(1,2),2))
design_mat = make_design(bio,batch, W = NULL)

log_expr = matrix(rnorm(20),ncol = 4)
adjusted_log_expr = lm_adjust(log_expr = log_expr,
  design_mat = design_mat,
  batch = batch)

YosefLab/scone documentation built on Oct. 21, 2024, 4:39 p.m.