Generating reports copenhagenweb.math.ku.dk/~richard/courses/bioconductor2007/CDrom/biomaRt/... · Microarray PhD Course November 2007 Example Genomic signatures to guide the use

Microarray PhD CourseNovember 2007

Generating reports

Steffen Durinck


Traditional way of writingreports

• Analyze the data• Copy results of the analysis in a report• Copy results from report into paper to

submit


Workflow


Report

• Easy to update analysis• Contain R analysis code• Dynamic• pdf format


Workflow


Report

• Easy to update analysis• Contain R analysis code• Dynamic• pdf format

Reproducable


Forensic bioinformatics

• Keith Baggerley recent talk coined term“Forensic Bioinformatics

What the methods section in the papersays that has been done

Try to reconstruct the analysis that wasactually performed


Example

Genomic signatures to guide the use ofchemotherapeutics

Potti et al., Nature Medicine - 12, 1294 - 1300 (2006)

Microarrays: retracing steps - pp1276 - 1277Kevin R Coombes, Jing Wang & Keith A BaggerlyThis week’s Nature Medicine


Example

• NCI60 cell lines– Drug response data– Microarray expression data

• Use NCI60 expression data to find set ofgenes that can be used as predictor fortreatment outcome in cancer tumors


Example

“Recent studies have suggested how this approach might workusing a widely-used panel of cell lines, the NCI60, to assemble theresponse signatures for several drugs. Unfortunately, ambiguitiesassociated with analyzing the data have made these resultsdifficult to reproduce.”

Keith Baggerly - Mathematical Systems Biology of Cancer II talk


Example

• Prediction on other samples workedusing their software

• But same gene set could also predictoutcome for other drugs????

Keith Baggerly - Mathematical Systems Biology ofCancer II talk


Forensic bioinformatics:investigation results

• Gene signature was offset by 1 -> indexingerror

• Training of algorithm was performed ontraining + test set=> Near perfect prediction results

Keith Baggerly - Mathematical Systems Biology ofCancer II talk


Sweave

• Developed by Friedrich Leisch• A tool that allows Embedding of R code

in LaTeX documents• Code, results and descriptions are

presented in a consistent way


Sweave

• Tables, graphs generated on the fly• Dynamic reports which can be updated

automatically


Sweave - how it works

Write mix of LaTeX code and R code in.Rnw file

Run Sweave function in R on the created.Rnw

Run latex on the created .tex file


Rnw example\documentclass[11pt]{article}\usepackage{hyperref}\usepackage{url}\usepackage[authoryear,round]{natbib}\newcommand{\Rfunction}[1]{{\texttt{#1}}}\newcommand{\Robject}[1]{{\texttt{#1}}}\newcommand{\Rpackage}[1]{{\textit{#1}}}

\author{Steffen Durinck\footnote{[email protected]}}\begin{document}\title{PhD course Copenhagen}\maketitle


Rnw example\section{Selecting a BioMart database and dataset}

\subsection{Selecting a BioMart database}

Every analysis with \Rpackage{biomaRt} starts with selecting aBioMart database to use. A first step is to check which BioMartweb services are available. The function \Rfunction{listMarts}will display all available BioMart web services

\begin{scriptsize}<<>>=library(biomaRt)listMarts()@\end{document}


Rnw example

> Sweave(“example.Rnw”)

% latex example.tex

example.pdf


Rnw example


Rnw

• R-code enclosed in

<<>>= R-code goes here@


Sweave options

• echo:• TRUE/FALSE• To display code or not

• eval:• TRUE/FALSE• To evaluate code or not


Sweave options

• fig:• TRUE• To plot and include images


Rnw

Lets create a data object<<>>= a = seq(1:4)@Lets create a second object<<echo = FALSE>>=b = a^2@


Rnw

Now some code we don’t want to be evaluated<<eval=FALSE>>= print(a)@Lets make a plot<<fig=TRUE>>=plot(a,b, pch=16, col=‘lightblue’)@


tablesDo the analysis<<>>=data(cats, package=“MASS”)lmout = lm(Hwt ~ Bwt * Sex , data = cats)@

Create a table<<results = tex>>=library(xtable)xtable(lmout)@


Report versions of usedpackages

• Output of analysis may depend on version ofpackages used.

• E.g. GO or annotation packages• Report package vesion by including

toLatex(sessionInfo())

in your report.


Stangle

• Extract R-code only from Rnw file

> Stangle("rnw-example.Rnw")Writing to file rnw-example.R

Code can then be sourced (and run) in R


Weaver

• Developed by Seth Falcon• Caching of results• Avoids having to re-compute a

computationally intensive piece of codeeach time one changes the document


Weaver

Usage:

In Rnw file:<<cache = TRUE>>

your code to cache @

Sweave(“yourRnw.Rnw”, driver = weaver())


Weaver

• Note:

– No side effects (e.g. printing, plotting) will work ina cache = TRUE code chunk

– If change in code chunk is detected cache will berecomputed

– Cache is not document specific. Two documentswith same piece of code in same directory will givesame cache!


Compendia

• Combine data and data analysis in an Rpackage

• Easy way to add to publication• Can also be published on BioC site as

example data analysis


Live demo


References

http://www.ci.tuwien.ac.at/~leisch/Sweave/

Generating reports copenhagenweb.math.ku.dk/~richard/courses/bioconductor2007/CDrom/biomaRt/... · Microarray PhD Course November 2007 Example Genomic signatures to guide the use

Documents