%
% NOTE -- ONLY EDIT THE .Rnw FILE!!!  The .tex file is
% likely to be overwritten.
%
\documentclass[12pt]{article}

\usepackage{amsmath,pstricks}
\usepackage[authoryear,round]{natbib}
\usepackage{hyperref}


\textwidth=6.2in
\textheight=8.5in
%\parskip=.3cm
\oddsidemargin=.1in
\evensidemargin=.1in
\headheight=-.3in

\newcommand{\scscst}{\scriptscriptstyle}
\newcommand{\scst}{\scriptstyle}


\newcommand{\Rfunction}[1]{{\texttt{#1}}}
\newcommand{\Robject}[1]{{\texttt{#1}}}
\newcommand{\Rpackage}[1]{{\textit{#1}}}
\newcommand{\Rmethod}[1]{{\texttt{#1}}}
\newcommand{\Rfunarg}[1]{{\texttt{#1}}}
\newcommand{\Rclass}[1]{{\textit{#1}}}

\textwidth=6.2in

\bibliographystyle{plainnat} 
 
\begin{document}
%\setkeys{Gin}{width=0.55\textwidth}



\subsection{Working with Neve2006}

We are going to examine some aCGH copy number data
as published by Neve et al 2006.

<<getdat>>=
library(Neve2006)
data(neveCGHmatch)
chr = function(x) pData(featureData(x))$Chrom
kb = function(x) pData(featureData(x))$kb
nc17 = neveCGHmatch[ chr(neveCGHmatch) == 17, ]
<<lk1,fig=TRUE>>=
par(mfrow=c(3,2))
for (i in 1:6)
 plot( logRatios(nc17)[,i] ~ kb(nc17), main=nc17$geneCluster[i],
  ylab="log ratio", xlab="on chr 17")
par(mfrow=c(1,1))
@

\subsection{Segmentation via rpart}

First we are going to use rpart as a device for estimating
a piecewise constant model for log ratios.  No optimality is
claimed for the technique; it can be compared to tilingArray::segment
and DNAcopy::segment.  However, there are infrastructure gains
obtained by using a bona fide modeling tool, which has a generic predict().

We develop additional infrastructure here:
<<build,results=hide>>=
library(rpart)
setOldClass("rpart")
setClass("rpSeg", representation(
  obj="rpart", samp="numeric", LR="numeric", KB="numeric",
  chr="character", rpCall="call", pdLabel="character"))
setMethod("show", "rpSeg", function(object) {
 cat("rpart segmentation for sample", object@samp, "phenoLabel", 
    object@pdLabel, "\n")
 cat("range of clone locations on chr", object@chr, "\n")
 print(range(object@KB))
})
setMethod("plot", "rpSeg", function(x, y, ...) {
 rng = range(x@KB)
 X = seq(rng[1], rng[2], 500) 
 Y = predict( x@obj, newdata=data.frame(KB=X) )
 plot( x@KB, x@LR, xlab=paste("kB on chr", x@chr), ylab="log ratio",
   main=paste("sample", x@samp, ";", x@pdLabel))
 points(X, Y, pch="-", cex=2)
})
treeSeg = function(es, samp, chr="17", pdv="geneCluster", ...) {
 LR = logRatios(es)[,samp]
 pdl = as.character(es[[pdv]][samp])
 KB = kb(es)
 ob = rpart(LR~KB, ...)
 new("rpSeg", obj=ob, samp=samp, LR=LR, KB=KB,
   chr=chr, rpCall=ob$call, pdLabel=pdl)
}
@

Now test it out:
<<lkts1>>=
ts1 = treeSeg(nc17, 1)
ts1
@
<<ss,fig=TRUE>>=
tsL = list()
for (i in 1:6) tsL[[i]] = treeSeg(nc17, i)
par(mfrow=c(3,2))
for (i in 1:6) plot(tsL[[i]])
par(mfrow=c(1,1))
@
 
\subsection{Open problems in integrative analysis}

\addtocounter{mlq}{1}
\bi
\item \textit{Question \MLQ.}  Add infrastructure to extract
estimated mean log ratio at selected chromosomal offsets for
all samples in a CGHset.
These extracted quantities 
can be regarded as new predictive features, and
will need names.  If possible,
write the infrastructure so that queries can be couched in
terms of gene symbols and extents.
\ei

\addtocounter{mlq}{1}
\bi
\item \textit{Question \MLQ.}  Is predictability of breast
cancer phenotype enhanced
when use is made of genomic aberration data in addition
to expression data?
\ei

\end{document}