fitModels: Fit a Poisson model for expression and read distribution...

In Senbee/Sequgio: Isoform expression estimation based on RNA-seq data

Description

Fit a joint statistical model that iteratively estimate non- uniform isoform-specific read distribution and gene isoform expression estimation.

Usage

fitModels(iGene, design, counts, probeLen = 50L, minoverlap=5L,robust = TRUE, use.joint = TRUE, verbose = FALSE, ls.start=FALSE,ridge.lambda = 0, maxit = 50L, std.err = FALSE, Q1 = 0.75, DF = 3L, fix.lambda = exp(20),max.exons = 100, use.trueiLen=FALSE, trueiLen, attr.iLen=FALSE, attr.df=FALSE,useC=FALSE)

Arguments

iGene	Charater. Gene IDs
design	A list constains the design matrix for every gene.
counts	A matrix contains the read counts.
probeLen	Integer. Probe length in base pair (bp). Default is set to50.
minoverlap	Integer. Minimum overlap on either side for junction reads
robust	Use robust estimation
use.joint	Logical. If TRUE, a joint statistical model is fitted, taking non-uniform read distribution into account. Default to TRUE
verbose	Logical. If TRUE, gene ID is printed after final estimation is done. Default to FALSE
ls.start	Logical. If FALSE, robust regression is performed. Default to TRUE
ridge.lambda	Penalization term
maxit	Integer. Maximum number of iterations in joint estimation of read distribution and expression level. Default is set to 50.
std.err	Logical. If TRUE, standard error of expression level is included in the output. Default to FALSE
Q1	Numeric. The percentile of residuals used as a threshold in assigning weight. Default is set to 0.75.
DF	Numeric. Degress of freedom of the smoothing. Default is set to 3.
fix.lambda	Numeric. A fixed large penalization value used when there are too many exons in a gene. Default is set to exp(20).
max.exons	Numeric. If the number of exons in a gene exceeds this number, a fixed large lambda value is used in smoothing read intensity. Default is set to 100.
use.trueiLen	Logical. If True, estimated true read distribution is employed, thus only expression level needs to be estimated. Default to FALSE
trueiLen	True read distribution
attr.iLen	Logical. If TRUE estimated read intensity is included in the output. Default to FALSE
attr.df	Logical. If TRUE degree of freedom used in smoothing read intensity is included in the output. Default to FALSE
useC	Logical. If TRUE C code would be called. Default to FALSE (Experimental and still under testing)

Value

A list contains expression leve, estimated read distribution, standard error of expression level and etc.

Author(s)

Stefano Calza <[e-mail]>, Chen Suo, Agus Salim and Yudi Pawitan

References

Suo C, Calza S, Salim A, Pawitan Y. Joint estimation of isoform expression and isoform-specific read distribution using multisample RNA-Seq data. Bioinformatics. 2014 Feb 15;30(4):506-13

See Also

makeXmatrix, getCounts

Examples

data(Counts)
data(Design)

iGenes <- names(Design)

fit1 <- fitModels(iGenes[21],Design,allCounts)
fit2 <- lapply(iGenes[21:22],fitModels,design=Design,counts=allCounts)

Senbee/Sequgio documentation built on May 9, 2019, 1:21 p.m.