% NOTE -- ONLY EDIT THE .Rnw FILE!!!  The .tex file is
% likely to be overwritten.
%% Modified using Sweave2knitr().

% \VignetteIndexEntry{Pigengene: Computing and using eigengenes}
%\VignetteDepends{Pigengene}
%\VignetteKeywords{Gene expression, Network, Biomedical Informatics, Systems Biology}
%\VignettePackage{Pigengene}
%\VignetteEngine{knitr::knitr}

\newcommand{\pipa}{\Biocpkg{Pigengene} }

\documentclass[12pt]{article}

<<style, eval=TRUE, echo=FALSE, results="asis">>=
BiocStyle::latex()
@

\bioctitle{\Biocpkg{Pigengene}: Computing and using eigengenes}
\author{Habil Zare}
\date{Modified: 26 April, 2016. Compiled: \today}

\begin{document}

\maketitle

\tableofcontents


\section{Introduction}
Gene expression profiling technologies such as microarray or RNA Seq provide 
valuable datasets, however, inferring biological information from these data 
remains cumbersome. \pipa address two challenges:
\begin{enumerate}
\item 
  \textbf{Curse of dimensionality:} The number of features in an expression profile is
  usually very high. For instance, there are about 20,000 genes in human. In contrast, the number
  of samples (patients) is often very limited in practice, and may not exceed a few
  hundreds. Yeung et al. have shown that standard data reduction methods such as principal 
  component analysis (PCA) are not appropriate to directly apply on gene expression data 
  \cite{yeung2001principal}.
  Instead, \pipa  addresses this challenge by applying PCA on gene modules.
\item
  \textbf{Normalization:} Data produced using different technologies, or in different labs,
  are not easily comparable. \pipa identifies {\em eigengenes}, informative biological signatures
  that are robust with respect to the profiling platform. For instance, it
  can identify the signatures (compute the eigengenes) on microarray data, and infer them
  on biologically-related RNA Seq data. The resulting  signatures are directly comparable even if the
  set of samples (patients)  are independent and disjoint in the two analyzed datasets.
\end{enumerate}


\section{How to run \pipa?}
\subsection{Installation}
\pipa is an \R{} package that can be downloaded and 
installed from \Bioconductor{} by the followig commands in R:
\newline\newline
\texttt{source("http://bioconductor.org/biocLite.R")}
\newline
\texttt{biocLite("Pigengene")}

Alternatively, if the source code is already available, 
the package can be installed by the following command in Linux:
\newline
$\\$
\texttt{R CMD INSTALL Pigengene\_x.y.z.tar.gz}
\newline
$\\$
where x.y.z. determines the version. The second approach requires all 
the dependencies be installed manually, therefore, the first approach
is preferred.

\subsection{A quick overview}
\pipa identifies gene modules (clusters), computes an eigengene for each 
module, and uses these biological signatures as features for classification.
The main function is \Rfunction{one.step.pigengene} which requires a gene 
expression profile and the corresponding conditions (types).
Individual functions are also provided to facilitate running the pipeline in a 
customized way. The inferred biological signatures (eigengenes)
are useful for supervised or unsupervised analyses.

\subsection{What is an eigengene?}
In most functions of this package, eigenegenes are computed or used as robust
biological signatures. Briefly, each eigengene is a weighted average of the
expression of all genes in a given set of genes (also known as a gene module or a cluster of genes).
The weights are adjusted  in a way that the explained variance is maximized. 
This guarantees that the loss in the biological information in minimized.

\subsection{A toy example}
For a quick start, the application of \pipa pipeline on some leukemia dataset is
demonstrated below \cite{mills2009microarray}. 
The first step is to load the package and data in \R{}:

<<loading, fig.width=6, fig.height=6, echo=TRUE>>=
library(Pigengene)
data(aml)
data(mds)
d1 <- rbind(aml,mds)
Labels <- c(rep("AML",nrow(aml)),rep("MDS",nrow(mds)))
names(Labels) <- rownames(d1)
Disease <- as.data.frame(Labels)
h1 <- pheatmap.type(d1[,1:20],annRow=Disease,show_rownames=FALSE)
@ 

Please note that the provided data in the package is
sub-sampled for a quicker demonstration. For real applications, the expression of
thousands of genes should be provided in order to the co-expression network analysis to 
be appropriate. It is common to first perform differential expression analysis, sort all the genes based
on p-values, and use the top-third as the input \cite{zhang2013integrated}.
Analyzing such input with \pipa can take a few hours and may require 5-10 GB of memory. 
The following command runs Pigengene pipeline on the toy data:
%%
<<oneStep, echo=TRUE>>=
p1 <- one.step.pigengene(Data=d1,saveDir='pigengene', bnNum=0, verbose=1,
      seed=1, Labels=Labels, toCompact=FALSE, doHeat=FALSE)
@ 

Results and figures are saved in \Rcode{pigengene} folder under the current directory.  
For more advanced applications, the user is encouraged to analyze the data step-by-step
and customize  the individual functions such as \Rfunction{compute.pigenegene} and 
\Rfunction{make.decision.tree}.

In addition to the provided decision trees, the user can also take alternative approaches to perform 
classification, clustering, survival analysis, etc. {\em using eigengenes as robust biological
signatures (informative features)}. Eigengenes and other useful objects can be retrieved from the output.
For instance, \Rcode{c5treeRes} is a list containing the results of fitting decision trees to
the eigengenes.  As shown above, a couple of trees were fitted, one per a value for 
\Rcode{minPerLeaf}. The following command plots the tree corresponding to 34, i.e.,
it was fitted requiring the minimum number of samples per every leaf to be at least 34.
%%
<<tree, fig.width=5, fig.height=5, echo=TRUE>>=
plot(p1$c5treeRes$c5Trees[["34"]])
@ 
The tree corresponding to other values are saved in \Rcode{pigengene} folder.
Of note, is the \Robject{pigenegene} object that contains the matrix of inferred eigenegenes.
Each row corresponds to a sample, and each column represents an eigengene.
%%
<<pigengene, fig.width=5, fig.height=5, echo=TRUE>>=
dim(p1$pigengene$eigengenes)
p1 <- pheatmap.type(p1$pigengene$eigengenes,annRow=Disease,show_rownames=FALSE)


@ 
 

\subsection{Citation}
The methodology and an interesting application of \pipa on studying
hematological malignancies is presented in the following reference \cite{foroushani2016large}.
<<citation, results='asis', eval=TRUE>>=
citation("Pigengene")
@ 

\section{Session Information}
The output of \Rfunction{sessionInfo} on the system that compiled 
this document is as follows:

<<sessionInfo, results='asis', eval=TRUE>>=
toLatex(sessionInfo())
@

\bibliography{pigengene}

\end{document}