estimate.rates: Wrapper function to estimate gene- and time specific...

In carinademel/RNAlife: Quantifying kinetic parameters to determine the RNA life cycle

Description

Wrapper function to estimate gene- and time specific concentrations for labeled (alpha) and unlabeled (beta) RNA for whole count table The initial values for alpha and beta can be provided, but are otherwise estimated from the data. In case the estimation yields negative or too small values they are replaced by alphabeta.min. For conditions, where no counts are available in both Labeled and Total samples, no rate estimation takes place and NA values are returned.

Usage

estimate.rates(counts, dispersion, lengths, conditions, conditions.labeling,
  replicates, cross.cont, seq.depths, lab.time, N = 1,
  consider.replicates = TRUE, rep = 1, logloglik = TRUE,
  alpha.initial = NULL, beta.initial = NULL, alphabeta.min = 1e-04,
  gene.indices = NULL, debug = FALSE)

Arguments

counts	count table, counts per gene (rows) and sample (columns)
dispersion	matrix or dataframe with gene-specific disperion estimates (alpha) for labeled and total sample to be used in Negative Binomial. Rownames should correspond to rownames in count matrix
lengths	gene lengths, should be in same order as rownames of count table
conditions	vector of time points/treatment of experiment for each sample
conditions.labeling	vector indicating for each sample if it was labeled RNA or total RNA
replicates	replicate number of each sample
cross.cont	vector with cross contamination rate values for each sample (total RNAseq cross.cont=1)
seq.depths	vector with sequencing depth values for each sample
lab.time	vector of same length as unique conditions, indicating labeling duration
N	number of cells
consider.replicates	should - if available - be replicates taken into account or not? If not, rep replicates is only considered for estimation
rep	if no replicate is available this is replicate that should be chosen
logloglik	should log-loglikelihood function be used?
alpha.initial	initial value for alpha, can be numeric (same for all genes and timepoints), a vector (same for all timepoints of one gene), or a matrix (individual for each gene and time point)
beta.initial	initial value for beta, can be numeric (same for all genes and timepoints), a vector (same for all timepoints of one gene), or a matrix (individual for each gene and time point)
alphabeta.min	minimum value as initial value for alpha and beta rates
gene.indices	indices of genes that should be used for calculation (in order to run code on smaller subset)
debug	should debugging modus be used?

Value

estimates for alpha and beta, expected counts in Labeled and Total and loss for each gene and each labeling time point

Author(s)

Carina Demel

Examples

data(gene.counts)
lengths = rnbinom(nrow(gene.counts), mu=1000, size=10)
data(spikein.counts)
data(spikein.labeling)
data(spikein.lengths)
spikeins = rownames(spikein.counts)
data(samples)
norm.res = spikein.normalization(spikein.counts, spikeins, spikein.lengths, spikein.labeling, 
	samples=colnames(spikein.counts), samples$conditions.labeling)
seq.depths = norm.res$sequencing.depth
cross.cont = norm.res$cross.contamination
dispersion = estimateGeneDispersion(gene.counts, samples$conditions.labeling, samples$conditions)
fittingres = estimateAlphaBeta.genes(1:12, gene.counts, dispersion, lengths, 
	samples$conditions, samples$conditions.labeling, samples$replicates, cross.cont, seq.depths,
 N=1, consider.replicates=TRUE, lab.time=c(5,5), alpha.initial=1, beta.initial=1)

carinademel/RNAlife documentation built on May 13, 2019, 12:43 p.m.