%
% NOTE -- ONLY EDIT THE .Rnw FILE!!!  The .tex file is
% likely to be overwritten.
%
%\VignetteIndexEntry{Annotation Overview}
%\VignetteDepends{Biobase, genefilter, annotate, hgu95av2.db}
%\VignetteKeywords{Expression Analysis, Annotation}
%\VignettePackage{annotate}

\documentclass{article}


\newcommand{\Rfunction}[1]{{\texttt{#1}}}
\newcommand{\Rmethod}[1]{{\texttt{#1}}}

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

\usepackage{hyperref}

\usepackage[authoryear,round]{natbib}
\usepackage{times}

\bibliographystyle{plainnat}

\begin{document}
\title{Bioconductor: Annotation Package Overview}
\maketitle{}

\section{Overview}

In its current state the basic purpose of \Rpackage{annotate} is to
supply interface routines that support user actions that rely on the
different meta-data packages provided through the Bioconductor
Project. There are currently four basic categories of functions that
are contained in \Rpackage{annotate}.
\begin{itemize}
\item Interface functions for getting data out of specific meta-data
      libraries.
\item Functions that support querying the different web services
      provided by the National Library of Medicine (NLM), and the
      National Center for Biotechnology Information (NCBI).
\item Functions that support organizing and structuring chromosomal
      location data to support some of the gene plotting and gene
      finding routines in \Rpackage{geneplotter}.
\item Functions that produce HTML output, hyperlinked to different web
      resources, from gene lists.
\end{itemize}

We will briefly describe the second and third of these different
aspects and then for the remainder of this vignette concentrate on the
first category. The other three have their own vignettes.


There are two different, but complementary strategies for accessing
meta-data. One is to use highly curated data that have been assembled
from many different sources and a second is to rely on on-line
sources. The former tends to be less current but more comprehensive
while the latter tends to be current but can be less comprehensive and
difficult to reproduce as the sources themselves are constantly
evolving. To address the second of these we develop and distribute
software that can take advantage of the web services that are
provided. Perhaps the richest source of these is provided by the
National Library of Medicine. Further details on accessing these
resources are provided in \cite{PubMedRnews} and \cite{PubMedVignette}.

While the chromosomal location is not always of interest, in certain
situations, especially the study of genetic diseases it is important
that we be able to associate particular genes with locations on
chromosomes. We provide a complete set of functions that map from
LocusLink identifiers to chromosomal location. Examples and further
discussion are provided in \cite{ChromLocVignette}.

Producing output that helps users navigate and understand the results
of an analysis is a very important aspect of any data analysis. Since
one of the primary tasks that is carried out when analysing gene
expression data is to create lists of interesting genes we have
provided some simple infrastructure that will help produce a
hyperlinked output page for any given set of genes. A more substantial
and comprehensive approach has be taken by C. Smith in the
\Rpackage{annaffy} package. A vignetted for using the functions
provided in the \Rfunction{annotate} package is provided in
\cite{HTMLVignette}.

We now turn our attention to interfaces to the meta-data packages and
how and when they will be useful.  The annotation library provides
interface support for the different meta-data packages
(\url{http://www.bioconductor.org/data/metaData.html} that are available
through the Bioconductor Project. We have tried to make these
different meta-data packages modular, in the sense that all of them
should have similar sets of mappings from manufacturer IDs to specific
biological data such as chromosomal location, GO, and LocusLink
identifiers.

Annotation in the Bioconductor Project is handled by two systems. One,
\Rpackage{AnnBuilder} is a system for assembling and relating the data from
various sources. It is much more {\em industrial} and takes advantage
of many different non-R tools such as Perl and XML. The second package
is {\em annotate}. This package is designed to provide experiment
level annotation resources and to help users associate the outputs of
\Rpackage{AnnBuilder} with their particular data.

There will be some need for the structure of the meta-data packages to
evolve over time and by making use of the functions provided in
\Rpackage{annotate} users and developers should shield themselves from
many of the changes. We will endeavor to keep the \Rpackage{annotate}
interfaces constant.

Any given experiment typically involves a set of known identifiers
(probes in the case of a microarray experiment). These identifiers are
typically unique (for any manufacturer). This holds true for any of
the standard databases such as LocusLink.
However, when the identifiers from one source are linked to the
identifiers from another there does not need to be a one--to--one
relationship. For example, several different Affymetrix accession
numbers correspond to a single LocusLink identifier. Thus, when going
one direction (Affymetrix to LocusLink) we have no problem, but when
going the other we need some mechanism for dealing with the
multiplicity of matches.

\subsection*{A Short Example}

We will consider the Affymetrix human gene chip, \texttt{hgu95av2},
for our example. We first load this chip's package and \Rpackage{annotate}.

<<loadlibs>>=
library("annotate")
library("hgu95av2.db")
ls("package:hgu95av2.db")
@
We see the listing of twenty or thirty different R objects in this
package. Most of them represent mappings from the identifiers on the
Affymetrix chip to the different biological resources and you can find
out more about them by using the R help system, since each has a
manual page that describes the data together with other information
such as where, when and what files were used to construct the
mappings. Also, each meta-data package has one object that has the
same name as the package basename, in this case it is \Robject{hgu95av2}. This
is function and it can be invoked to find out some of the different
statistics regarding the mappings that were done. Its manual page
lists all data resources that were used to create the meta-data package.

<<qc>>=
hgu95av2()
@

Now let's consider a specific object, say the
\Robject{hgu95av2ENTREZID} object.
<<locusid>>=
hgu95av2ENTREZID
@
If we type its name we see that it is an R \texttt{environment}; all
this means is that it is a special data type that is efficient at
storing and retrieving mappings between symbols (the Affymetrix
identifiers) and associated values (the LocusLink IDs).

We can retrieve values from this environment in many different ways
(and the reader is referred to the R help pages for more details on
using some of the functions described briefly here).
Suppose that we are interested in finding the LocusLink ID for the
Affymetrix probe, \texttt{1000\_at}. Then we can do it in any one of
the following ways:
<<getting>>=

get("1000_at", env=hgu95av2ENTREZID)
hgu95av2ENTREZID[["1000_at"]]
hgu95av2ENTREZID$"1000_at"

@
And in all cases we see that the LocusLink identifier is
\Sexpr{hgu95av2ENTREZID[["1000_at"]]}.

If you want to get more than one
object from an environment you also have a number of choices. You can
extract them one at a time using either a \texttt{for} loop or
\Rfunction{apply} or \Rfunction{eapply}. It will be more efficient to
use \Rfunction{mget} since it does the retrieval using internal code
and is optimized. You may also want to turn the environment into a
named list, so that you can perform different list operations on it,
this can be done using the function \Rfunction{contents} or
\Rfunction{as.list}.

<<tolist>>=
LLs = as.list(hgu95av2ENTREZID)
length(LLs)
names(LLs)[1:10]

@

\section{Developers Tips}

Software developers that are building other tools that might make use
of the infrastructure produced here should make use of the
\Rfunction{get*} family of functions. Examples include
\Rfunction{getEG}, \Rfunction{getGO} and so on. There are two reasons
for using these functions.

First, they allow you to implement functionality that is independent
of the meta-data packages. Since each of these functions takes two
arguments, one a list of the manufacturers ids and the second the
\textit{basename} of the annotation package these functions will
provide the correct results for all annotation packages.

A second reason to make use of these interface functions is that they
are guaranteed not to change. The underlying structure of the
meta-data packages may change and developers that access this directly
will find that their code needs to be updated regularly. But
developers that make use of these interface functions will find that
their code needs much less maintenance.

\section{Session Information}

The version number of R and packages loaded for generating the vignette were:

<<echo=FALSE>>=
sessionInfo()
@

\bibliography{annotate}

\end{document}