fastseg (version 1.18.0)

fastseg: Detection of breakpoints using a fast segmentation algorithm based on the cyber t-test.

Description

Detection of breakpoints using a fast segmentation algorithm based on the cyber t-test.

Usage

fastseg(x, type = 1, alpha = 0.05, segMedianT, minSeg = 4, eps = 0, delta = 5, maxInt = 40, squashing = 0, cyberWeight = 10)

Arguments

x

Values to be segmented either in the format of a sorted GRanges object, ExpressionSet object, matrix or vector.

type

Parameter that sets the type of test. If set to 1 a test of the left against the right window is performend. If set to 2 the segment is also tested against the global mean. (Default = 1).

alpha

A value between 0 and 1 is interpreted as the ratio of initial breakpoints. An integer greater than one is interpreted as number of desired breakpoints. Increasing this parameter leads to more segments. (Default = 0.1)

segMedianT

A numeric vector of length two with the thresholds of segments' median values that are considered as significant. Only segments with a median above the first or below the second value are kept in a final merging step. If missing the algorithm will try to find a reasonable value by using z-scores. (Default "missing".)

minSeg

The minimal segment length. (Default = 4).

eps

Minimal distance between consecutive values. Only consecutive values with a minimium distance of "eps" are tested. This makes the segmentation algorithm even faster. If all values should be tested "eps" can be set to zero. If missing the algorithm will try to find a reasonable value by using quantiles. (Default = 0.)

delta

Segment extension parameter. If delta consecutive extensions of the left and the right segment do not lead to a better p-value the testing is stopped. (Default = 5).

maxInt

Maximal length of the left and the right segment. (Default = 40).

squashing

The degree of squashing of the input values. If set to zero no squashing is performed. (Default = 0).

cyberWeight

The nu parameter of the cyber t-test. Can be interpreted as the weight of the global variance. The higher the value the more small segments with high variance will be significant. (Default = 10).

...

Further arguments passed to the plot function.

Value

A data frame containing the segments.

Examples

library(fastseg)

#####################################################################
### the data
#####################################################################
data(coriell)
head(coriell)

samplenames <- colnames(coriell)[4:5]
data <- as.matrix(coriell[4:5])
data[is.na(data)] <- median(data, na.rm=TRUE)
chrom <- coriell$Chromosome
maploc <- coriell$Position


###########################################################
## GRanges
###########################################################

library("GenomicRanges")

## with both individuals
gr <- GRanges(seqnames=chrom,
        ranges=IRanges(maploc, end=maploc))
mcols(gr) <- data
colnames(mcols(gr)) <- samplenames
res <- fastseg(gr)

## with one individual
gr2 <- gr
data2 <- as.matrix(data[, 1])
colnames(data2) <- "sample1"
mcols(gr2) <- data2
res <- fastseg(gr2)


###########################################################
## vector
###########################################################
data2 <- data[, 1]
res <- fastseg(data2)



###########################################################
## matrix
###########################################################
data2 <- data[1:400, ]
res <- fastseg(data2)