%\VignetteIndexEntry{tigre Quick Guide}
%\VignetteKeywords{TimeCourse, GeneExpression, Transcription}
%\VignettePackage{tigre}
\documentclass[a4paper]{article}
\usepackage{url}
\usepackage[authoryear,round]{natbib}

\title{tigre Quick Guide}
\author{Antti Honkela, Pei Gao, Jonatan Ropponen, \\
  Magnus Rattray, and Neil D. Lawrence}

\newcommand{\Rfunction}[1]{{\texttt{#1}}}
\newcommand{\Robject}[1]{{\texttt{#1}}}
\newcommand{\Rpackage}[1]{{\textit{#1}}}
\newcommand{\tigre}{\Rpackage{tigre}}

\begin{document}
\maketitle
\SweaveOpts{keep.source=TRUE}

\section{Abstract}

The \tigre{} package implements our methodology of Gaussian process
differential equation models for analysis of gene expression time
series from single input motif networks.  The package can be used for
inferring unobserved transcription factor (TF) protein concentrations
from expression measurements of known target genes, or for ranking
candidate targets of a TF.

The purpose of this quick guide is to present a small subset of
examples from the User Guide that can be run very quickly.  For a more
comprehensive (although slower-running) presentation, please refer to
the \tigre{} User Guide.

\section{Citing \tigre{}}

The \tigre{} package is based on a body of methodological research.
Citing \tigre{} in publications will usually involve citing one or
more of the methodology papers
\citep{Honkela2010PNAS,Gao2008,Lawrence2007} that the software is
based on as well as citing the software package itself
\citep{Honkela2011}.

<<echo=FALSE, eval=TRUE>>=
options(width = 60)
@

\section{Introductory example analysis - Drosophila development}
\label{section:Introductory example}

In this section we introduce the main functions of the \Rpackage{puma}
package by repeating some of the analysis from the PNAS
paper~\citep{Honkela2010PNAS}\footnote{Note that the results reported
  in the paper were run using an earlier version of this package for
  MATLAB, so there can be minor differences.}.

\subsection{Installing the \tigre{} package}

The recommended way to install \tigre{} is to use the
\Rfunction{biocLite} function available from the bioconductor
website. Installing in this way should ensure that all appropriate
dependencies are met.

<< eval=FALSE >>==
source("http://www.bioconductor.org/biocLite.R")
biocLite("tigre")
@

% To install the tigre software, unpack the software and run
% \begin{verbatim}
% R CMD INSTALL tigre
% \end{verbatim}

To load the package start R and run
<<>>=
library(tigre)
@ 

\subsection{Loading the data}

To get started, you need some preprocessed time series expression
data.  If the data originates from Affymetrix arrays, we highly
recommend processing it with \Rfunction{mmgmos} from the
\Rpackage{puma} package.  This processing extracts error bars on the
expression measurements directly from the array data to allow judging
the reliability of individual measurements.  This information is
directly utilised by all the models in this package.

To start from scratch on Affymetrix data, the .CEL files from
\url{ftp://ftp.fruitfly.org/pub/embryo_tc_array_data/} may be
processed using:
<<eval=FALSE>>=
# Names of CEL files
expfiles <- c(paste("embryo_tc_4_", 1:12, ".CEL", sep=""),
              paste("embryo_tc_6_", 1:12, ".CEL", sep=""),
              paste("embryo_tc_8_", 1:12, ".CEL", sep=""))
# Load the CEL files
expdata <- ReadAffy(filenames=expfiles,
                    celfile.path="embryo_tc_array_data")
# Setup experimental data (observation times)
pData(expdata) <- data.frame("time.h" = rep(1:12, 3),
                             row.names=rownames(pData(expdata)))
# Run mmgMOS processing (requires several minutes to complete)
drosophila_mmgmos_exprs <- mmgmos(expdata)
drosophila_mmgmos_fragment <- drosophila_mmgmos_exprs
@ 

This data needs to be further processed to make it suitable for our
models.  This can be done using
<<eval=FALSE>>=
drosophila_gpsim_fragment <-
  processData(drosophila_mmgmos_fragment,
              experiments=rep(1:3, each=12))
@ 

Here the last argument specifies that we have three independent time
series of measurements.

In order to save time with the demos, a part of the result of this is
included in this package and can be loaded using
<<>>=
data(drosophila_gpsim_fragment)
@ 

\subsection{Learning an individual model}

Let us now learn a model for the TF twist and one of its
potential targets
<<>>=
# The probe identifier for TF 'twi'
twi <- "143396_at"
# The probe identifier for the target gene
target <- "152715_at"

# Learn the model using only one of the 3 repeats in the data
model <- GPLearn(drosophila_gpsim_fragment[,1:12],
                 TF=twi, targets=target,
                 useGpdisim=TRUE, quiet=TRUE)

# Display the model parameters
show(model)
@ 

\subsection{Visualising the model}

The model can be plotted using the command
<<eval=FALSE>>=
GPPlot(model)
@ 
the results of which can be seen in Fig.~\ref{fig:model}.

\begin{figure}
  \begin{center}
<<fig=TRUE, echo=FALSE>>=
GPPlot(model)
@
\end{center}
\caption{Single target models for the gene FBgn0003486. The models for
  each repeated time series are shown in different columns.\label{fig:model}}
\end{figure}

\subsection{Ranking the targets}

Please refer to the User Guide for details on bulk ranking of
candidate targets.

\section{Session Info}

<<sessionInfo>>=
sessionInfo()
@ 

\bibliographystyle{plainnat}
\bibliography{gpsim}

\end{document}