Copy number profiles and complete 0/1 breakpoint annotations

Share:

Description

This data set contains two motivating segmentation problems for the SegAnnot algorithm. As described in the vignette, maximum-likelihood segmentations can be calculated using cghseg::segmeanCO, but none of the segmentations are consistent with the given breakpoint annotations. So SegAnnot can be used in these cases to find the best consistent segmentation in terms of the square loss.

Usage

1

Format

A list of two separate segmentation problems, named low.resolution and high.resolution. Each problem is a list of two data.frames: probes and annotations. The probes come from a microarray experiment, and the annotations come from an expert's interpretation of a scatterplot of the probe logratio versus position.

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
data(profiles,package="SegAnnot")
pro <- profiles$low
library(ggplot2)
p <- ggplot()+
  geom_tallrect(aes(xmin=min/1e6,xmax=max/1e6,fill=annotation),
                data=pro$ann)+
  geom_point(aes(position/1e6,logratio),data=pro$pro)+
  scale_fill_manual(values=breakpoint.colors)
print(p)
fit <- SegAnnotBases(pro$pro$log,pro$pro$pos,pro$ann$min,pro$ann$max)
pfit <- p+
  geom_segment(aes(first.base/1e6,mean,xend=last.base/1e6,yend=mean),
    data=fit$seg.df,colour=signal.colors[["estimate"]],lwd=2)+
  geom_vline(aes(xintercept=base/1e6),data=fit$break.df,
    colour=signal.colors[["estimate"]],lwd=2,linetype="dashed")
print(pfit)