seqCNA (version 1.18.0)

runSeqnorm: This function calls seqnorm, which normalizes the tumoural profile with respect to either a paired sample or GC content.

Description

The seqnorm method performs several regressions on homogeneous regions of the tumoural genomic profile.

Usage

runSeqnorm(rco, norm.win = NULL, method = "quadratic", lambdabreak=8, minSeg=7, maxSeg=35, nproc = 2, plots = TRUE, folder = NULL)

Arguments

rco

A SeqCNAInfo-class object, with read count (RC) and genomic information, normally the output of the readSeqsumm function.

norm.win

The genomic window size, given in kilobases, in which to normalize the read count profile. Leave it to the default NULL value for automatic selection. The window size should be the same as or a multiple of the summarization window in runSeqsumm. If not a multiple, the window size will be adjusted to the nearest one. For efficiency, we recommend to use the same value as the summarization window size, unless it is very small (e.g. 2Kbp).

method

A string being "loess", "cubic" or "quadratic", indicating which form of regression to use in the normalization. LOESS is more sensitive to noise and slower than the other two, which apply polynomial regression. In principle, we recommend using a quadratic polynomial regression, but there might be ocassions in which the others yield better fits.

lambdabreak

The lambda prime parameter in GLAD, indicating the penalty term used during the optimization of the number of breakpoints.

minSeg

The minimum number of segments in which the second pass regression is based. Too few would result in noisy estimates of the regression curve. We recommend to leave it to the default.

maxSeg

The minimum number of segments in which the second pass regression is based. Too many would include segments whose regression curve does not approximate the overall curve. Additionally, they would add additional execution time. We recommend to leave it to the default.

nproc

A value indicating how many processing cores to use in the segmentation prior to regression and in the segment regressions. Greater values speed up normalization using more CPU cores and RAM memory, but you should not use values greater than the number of cores in your machine. If unsure, the safest value is 1, but most computers nowadays are multi-core, so you could probably go up to 2, 4 or 8.

plots

A boolean value indicating whether to output normalization plots to the folder. These are useful for checking the correctness of the normalization and the improvements over the standard normalization.

folder

If the plots parameter is set to TRUE, path to the folder where the plots with normalization information are to be generated. If no folder is indicated or does not exist, plots will be displayed within R.

Value

A SeqCNAInfo-class object, with additional information on the normalized profile.

References

Analysis of array CGH data: from signal ratio to gain and loss of DNA regions. Hupe P, Stransky N, Thiery JP, Radvanyi F, Barillot E. Bioinformatics. 2004 Dec 12; 20(18):3413-22.

Examples

data(seqsumm_HCC1143)
rco = readSeqsumm(tumour.data=seqsumm_HCC1143)
rco = applyFilters(rco, 0, 1, 0, 2, FALSE, plots=FALSE)

### NORMALIZATION ###

rco = runSeqnorm(rco, plots=FALSE)