logVarGenes: Find highly variable genes (log linear model)

In wjawaid/bglab: Berthold Gottgens lab scRNA seq

Description

Find highly variable genes (log linear model)

Usage

logVarGenes(scd, minMean = 0, fraction = 0.05, lower = FALSE,
  residualsInLogSpace = TRUE, quadratic = TRUE, se = qnorm(p = 0.975))

Arguments

scd	Single Cell Dataset
minMean	Fit model to genes with mean expression gretaer than minMean
fraction	The fraction of cells on
lower	By default this function will only winsorise the upper end. If you also require the bottom end to be winsorised set this to TRUE.
residualsInLogSpace	Use lm instead of nls to optimise residuals in log-space.
quadratic	Fit an order 2 model to log cv-squared against log mean
se	The number of standard errors to return.

Details

Find highly variable genes (log linear model). Model is fitted to:

\frac{σ^{2}}{μ^{2}} = a * μ^{k}

Using the non-linear squares method, if residualsInLogSpace is FALSE. Otherwise using a linear model +/- a quadartic term.

Value

Returns a list of means, cv2, fit object and variable genes

Author(s)

Wajid Jawaid

wjawaid/bglab documentation built on May 4, 2019, 6:33 a.m.