Content uploaded by Thierry Lecroq
Author content
All content in this area was uploaded by Thierry Lecroq on Oct 05, 2015
Content may be subject to copyright.
2014, pages 1–2
BIOINFORMATICS APPLICATIONS NOTE doi:10.1093/bioinformatics/btu185
Genome analysis Advance Access publication April 9, 2014
Rgb: a scriptable genome browser for R
Sylvain Mareschal
1,2,3,
*, Sydney Dubois
1,2,3
, Thierry Lecroq
2,3,4
and Fabrice Jardin
1,2,3,
*
1
Centre Henri Becquerel, INSERM UMR 918, 76038 Rouen Cedex 1, France,
2
Normandy University, University of
Rouen, 76821 Mont-Saint-Aignan, France,
3
Institute for Research and Innovation in Biomedicine (IRIB), Haute-
Normandie, 76183 Rouen Cedex, France and
4
LITIS, INSA EA 4108, 76801 Saint-Etienne-du-Rouvray, France
Associate Editor: Inanc Birol
ABSTRACT
Summary: Thanks to its free licensing and the development of initia-
tives like Bioconductor, R has become an essential part of the bio-
informatics toolbox in the past years and is more and more confronted
with genomically located data. While separate solutions are available
to manipulate and visualize such data, no R package currently offers
the efficiency required for computationally intensive tasks such as
interactive genome browsing. The package proposed here fulfills this
specific need, providing a multilevel interface suitable for most needs,
from a completely interfaced genome browser to low-level classes
and methods. Its time and memory efficiency have been challenged
in a human dataset, where it outperformed existing solutions by sev-
eral orders of magnitude.
Availability and implementation: R sources and packages are freely
available at the CRAN repository and dedicated Web site: http://bio-
informatics.ovsa.fr/Rgb. Distributed under the GPL 3 license, compat-
ible with most operating systems (Windows, Linux, Mac OS) and
architectures.
Contact: maressyl@gmail.com or fabrice.jardin@chb.unicancer.fr
Supplementary information: Supplementary data are available at
Bioinformatics online.
Received on December 20, 2013; revised on March 3, 2014; accepted
on April 2, 2014
1 INTRODUCTION
The growing demand from the biology community for statistic-
ally robust approaches has made the R statistics-oriented script-
ing language an essential part of the bioinformatics toolbox. Its
graphical capabilities make it a valuable tool to produce publi-
cation-grade complex figures, whereas its computational effi-
ciency allows it to handle huge datasets, as currently required
in fields like transcriptomics or next-generation sequencing.
These qualities come with an open-source licensing and various
operating system ports that make it available virtually every-
where. Thanks to the Bioconductor initiative (Gentleman
et al., 2004), large amount of software is freely available as R
packages for tasks as diverse as microarray processing (Smyth,
2005;Venkatraman and Olshen, 2007), feature annotation
(Zhang et al., 2003)orsequenceanalysis(Anders and Huber,
2010).
Much of this software generates genomic data, i.e. lists of
chromosome regions defined by starting and ending coordinates.
Such data are usually subset using chromosomal coordinates
rather than row indexes, a paradigm R was not developed to
deal with. Bioconductor historically addressed this issue with
the RangedData class from the IRanges package, handling gen-
omic regions as ranges of integers (base positions). Its flexibility
and efficiency were extended a few years later by the
GenomicRanges package, making direct use of IRanges compo-
nents for the subsetting. For visualization, Bioconductor pro-
vides two solutions: rtracklayer and GViz. The former sends
data to visualize to the UCSC web genome browser (Kent
et al., 2002), a model that implies frequent comings and goings
between programs and consequent network burden. The latter is
more integrated within Bioconductor and produces static graph-
ics from the classes described above, in delays incompatible with
user interactivity and intensive computing.
The package described here reconciles these two aspects in a
coherent way, thus offering an interactive interface responsive
enough for comfortable browsing and atomic operations suitable
for computer-intensive algorithms.
2 IMPLEMENTATION
Rgb is implemented as an R package, providing various classes
and methods for scripts. It makes use of the Reference class
system, which offers an object-oriented framework similar to
what can be found in other languages such as Java or Cþþ.It
is flexible enough to suit the needs of the four categories of R
users:
R beginners will find in Rgb a complete graphical user inter-
face (GUI) allowing them to convert and visualize genomic
data without any command line. Stand-alone builds of Rgb
can even make the R dependency totally transparent.
Script writers, needing an exclusive Command Line
Interface to automate analysis, will find in Rgb classes and
methods able to handle their genomic data and produce
high-quality graphics from them.
Console users who use R as an exploratory tool without a
special need for reproducibility, will find Rgb’s ability to mix
the two interfaces particularly time saving.
Package developers may be interested in extending classes to
handle new data storage modes or representations, a process
greatly facilitated by the object-oriented design of Rgb.
The core of the subsetting system consists of a collection of C
functions making direct use of R libraries, and the graphical
interface relies on Tcl-tk. Both of them are natively managed
*To whom correspondence should be addressed.
ßThe Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com 1
Bioinformatics Advance Access published April 21, 2014
by guest on August 20, 2015http://bioinformatics.oxfordjournals.org/Downloaded from
by R, and have been successfully compiled and tested on
Windows and Linux operating systems. BAM file support can
be added by the Rsamtools package, which is part of
Bioconductor.
3 PERFORMANCES
To assert Rgb’s track.table class suitability for computer-inten-
sive tasks such as responsive genome browsing or overlap com-
putation, its time and maximal memory consumption on usual
atomic tasks were compared with existing solutions: standard R
data.frame, the more efficient data.table, IRanges’ RangedData,
GenomicRanges’ GRanges and GViz’s AnnotationTrack.Three
tasks were monitored as the most common atomic operations in
such data: genomic extraction by chromosomal coordinates, ex-
traction and modification of consecutive rows by indexes. The
291 128 exons recorded in the CCDS database (Pruitt et al.,
2009) for the human genome were used as a common dataset.
Benchmarking was performed on a mid-range desktop computer
(3.1 GHz Intel i3-2100 processor with 8 GB RAM) running R
3.0.2 in a 64-bit Fedora 18 distribution. Each measure was
made with R functions proc.time and gc in a fresh session, to
normalize garbage collection effects (scripts and dataset available
as Supplementary Data).
With the genomic extraction task (Fig. 1), GViz and
RangedData proved their unsuitability to intensive tasks, with
computing times nearing the second for a single extraction (typ-
ical genome browser representations contain several tracks to be
subset), and extravagant memory usage (hundreds of Mio for a
dataset of 30 Mio). GRanges and generic R solutions perform
better, but are still outperformed by Rgb by a magnitude of 50–
1500 on small genomic extractions.
Similar results can be observed with the modification task,
with Rgb outperforming R generic classes by a factor of 100 in
time consumption and 10 in memory usage. It still outperforms
bioinformatic solutions with the last task, but is overtaken here
by generic solutions such as data.frame and data.table. However,
as this kind of extraction is far less justified in a genomic context
than the previous one, this average performance can be accepted.
4CONCLUSION
Rgb provides several entry points to a consistent genome brows-
ing system, which may prove equally useful to users looking for
an interactive genome browser able to handle their R datasets
and developers needing efficient genomic subsetting capabilities
for their scripts. Its object-oriented paradigm and open-source
licensing make it easily extendable, and its performances over
existing solutions have been proven in a real genomic dataset.
ACKNOWLEDGEMENTS
The authors thank Christian Bastard for his helpful comments
and feedback during Rgb development and the many CGH-
array datasets he provided for testing.
Funding: This work was supported by a PhD grant from the
Re
´gion Haute-Normandie (France).
Conflict of Interest: none declared.
REFERENCES
Anders,S. and Huber,W. (2010) Differential expression analysis for sequence count
data. Genome Biol.,11,R106.
Gentleman,R.C. et al. (2004) Bioconductor: open software development for com-
putational biology and bioinformatics. Genome Biol.,5,R80.
Kent,W.J. et al. (2002) The human genome browser at UCSC. Genome Res.,12,
996–1006.
Pruitt,K.D. et al. (2009) The consensus coding sequence (CCDS) project: identifying
a common protein-coding gene set for the human and mouse genomes. Genome
Res.,19, 1316–1323.
Smyth,G.K. (2005) limma: linear models for microarray data. In: Gentleman,R.
et al. (ed.) Bioinformatics and Computational Biology Solutions Using R and
Bioconductor, Statistics for Biology and Health. Springer, New York, NY,
pp. 397–420.
Venkatraman,E.S. and Olshen,A.B. (2007) A faster circular binary segmentation
algorithm for the analysis of array CGH data. Bioinformatics,23, 657–663.
Zhang,J. et al. (2003) An extensible application for assembling annotation for gen-
omic data. Bioinformatis,19,155.
Fig. 1. Performance comparison of Rgb’s track.table class with existing
software. Three atomic tasks were benchmarked on a common dataset, in
terms of maximal memory usage and computing time, as recorded by R
2
S.Mareschal et al.
by guest on August 20, 2015http://bioinformatics.oxfordjournals.org/Downloaded from
