- TECHNICAL NOTE
- Open Access
- Open Peer Review
Colib’read on galaxy: a tools suite dedicated to biological information extraction from raw NGS reads
- Yvan Le Bras1Email author,
- Olivier Collin1,
- Cyril Monjeaud1,
- Vincent Lacroix2,
- Éric Rivals3,
- Claire Lemaitre4,
- Vincent Miele2,
- Gustavo Sacomoto2,
- Camille Marchet2,
- Bastien Cazaux3,
- Amal Zine El Aabidine3,
- Leena Salmela5,
- Susete Alves-Carvalho4,
- Alexan Andrieux4,
- Raluca Uricaru6, 7 and
- Pierre Peterlongo4
© Le Bras et al. 2016
- Received: 25 November 2014
- Accepted: 7 December 2015
- Published: 11 February 2016
With next-generation sequencing (NGS) technologies, the life sciences face a deluge of raw data. Classical analysis processes for such data often begin with an assembly step, needing large amounts of computing resources, and potentially removing or modifying parts of the biological information contained in the data. Our approach proposes to focus directly on biological questions, by considering raw unassembled NGS data, through a suite of six command-line tools.
Dedicated to ‘whole-genome assembly-free’ treatments, the Colib’read tools suite uses optimized algorithms for various analyses of NGS datasets, such as variant calling or read set comparisons. Based on the use of a de Bruijn graph and bloom filter, such analyses can be performed in a few hours, using small amounts of memory. Applications using real data demonstrate the good accuracy of these tools compared to classical approaches. To facilitate data analysis and tools dissemination, we developed Galaxy tools and tool shed repositories.
With the Colib’read Galaxy tools suite, we enable a broad range of life scientists to analyze raw NGS data. More importantly, our approach allows the maximum biological information to be retained in the data, and uses a very low memory footprint.
- Variant calling
- Whole-genome assembly-less treatment
- De Bruijn graph
- Bloom filter
- long read correction
For some years now, owing to the impact of high-throughput sequencing, also known as next-generation sequencing (NGS), genomics is witnessing profound changes. NGS technologies generate huge volumes of data, up to terabyte scale, and new types of raw and processed data. Usually, a generic assembly (preprocessing) is first applied to the raw sequences, and then, in a second step, the information of interest is extracted. This protocol may lead to a significant loss of information, or may generate chimerical results because of the heuristics used in the assembly. To circumvent this problem, we developed a set of innovative methods for extracting information of biological interest directly from NGS data, which allows the user to bypass a costly and often inaccurate assembly phase. Notably, the approaches developed do not require the availability of a reference genome. This considerably broadens the spectrum of applications. In this paper we present our approach, which relies on a tools suite born from the Colib’read  project and is dedicated to whole-genome assembly-free treatments. A set of six tools based on this framework, KISSPLICE , MAPSEMBLER2 , DISCOSNP , TAKEABREAK , COMMET , and LORDEC , are described below. To facilitate the use and the dissemination of our approach, we have developed Galaxy [8–11] tools and created repositories on GUGGO Tool Shed [12, 13]. We first highlight the fundamental computational concepts shared by the tools, and this is followed by the algorithmic developments and tool descriptions. Next, several applications using biological data are described and the Galaxy integration and dissemination processes are then detailed. Finally, the Galaxy integration and processes are described.
Summary of the Colib’read tools inputs and outputs
One or more RNA-seq read set(s)
SNPs, small indels, alternative splicing events
One or more raw genomic read set(s)
SNP sequences with their coverages
One or more raw genomic read set(s)
Pieces of known sequences, and associated raw read sets
Validation and visualization of genome structure near a locus of interest
Several raw metagenomic complex read sets
Global comparison of input sets at the read level
Illumina and PacBio read sets
Corrected PacBio read set
LORDEC uses short reads for correcting third-generation long reads, and finally COMMET (COmpare Multiple METagenomes) is dedicated to the comparison of numerous metagenomic read sets. Special care was given to limit both the memory and time requirements of all tools. Thus, five of the six tools are based on the usage of a compact representation of a de Bruijn graph, as explained in the next section.
Note that all the algorithms presented here were developed bearing in mind the need for simple and user-friendly tools. They can be applied on raw sequenced short reads, without requiring any pretreatment. However, if users are aware of bias such as contaminants or systematic sequencing errors, they can be used on preprocessed datasets and this can give better results.
A common kernel: the de Bruijn graph
Through the last decade, dBGs have been used extensively in the short read assembly framework. Indeed, the construction of such graphs is fast as it avoids any alignment computation, and it is memory efficient, as it compresses the read redundancy. In addition, since every nucleotide is explicitly present in this structure, sequence variants correspond to recognizable patterns. Therefore dBGs are well tailored for developing methods for detecting sequence polymorphism. DISCOSNP detects patterns generated by single-nucleotide polymorphisms (SNPs); KISSPLICE deals with RNA-seq data and finds patterns generated by SNPs, indels, and alternative splicing (AS) events; and TAKEABREAK detects patterns generated by inversions.
MAPSEMBLER2 and LORDEC are also based on a dBG, respectively for building a targeted assembly and as a reference for correcting third-generation long reads.
With the exception of KISSPLICE and COMMET, all the algorithms presented are based on an efficient representation of the de Bruijn graph with optimized bloom filters, implemented in the GATB C++ library , as used for the first time in the MINIA [15, 16] assembler.
A bloom filter is a probabilistic data structure that stores the presence/absence of items. It consists of a simple bit vector, initially all set to ‘0’. Any item is associated with a set of a few addresses in this vector (typically seven addresses). While adding an item, the corresponding bits are set to ‘1’. Note that a bit may be set to ‘1’ from several distinct items. While querying an item, if all its bits are equal to ‘1’ then the item is considered as indexed in the bloom filter. Conversely, if any of its bits are equal to ‘0’, the item is absent from the indexed data. The main advantage of the bloom filter is its simplicity and its low memory footprint. Its main disadvantage is that it is probabilistic: if the bloom filter answers ‘yes’ while querying the presence of an item, this answer may be wrong (with a controlled percentage).
Thus, the bloom filter representation has the main advantage of a very low memory footprint. For instance, nearly 3 billion reads (100 bp) were analyzed by DISCOSNP, using at most 5.7 GB of memory. Moreover, the low memory footprint does not imply a degradation in the running time. The COMMET tool, being a heuristic based only on a bloom filter, is also fast and has an extremely low memory footprint. Unfortunately the GATB  data structure does not yet allow the assignment of additional information to dBG nodes. In the KISSPLICE case, as presented in more detail in the following section, the nodes of the graph need to be tagged, thus requiring the use of an explicit dBG representation. Even if this representation is more resource intensive, it scales up perfectly in RNA-seq data problems.
Description of tools
DISCOSNP  is a reference-free SNP calling program that focuses on the detection of both heterozygous and homozygous isolated SNPs, from any number of sequencing datasets.
The DISCOSNP method rests on the following observation: in the dBG, a SNP generates a pair of paths composed of k vertices, which represent 2k-1 length sequences that are polymorphic on one position. This corresponds to a so-called bubble in the dBG, as depicted in Fig. 2. The model formalism can be found in .
DISCOSNP is composed of two modules: KISSNP2, followed by KISSREADS. The tool takes as input any number (potentially one) of read sets, i.e. samples. It processes all read sets together (creating the dBG and detecting the SNP-specific motifs) and outputs all isolated SNPs (for a given k) shared by any number of samples. The KISSNP2 output is a multi-FASTA file in which every consecutive pair of sequences corresponds to the two paths of a SNP (2k-1 sequences) together with their left and right contigs, which are reconstructed with the MINIA assembler . The KISSREADS module maps back input reads on the sequences of the predicted SNPs in order to validate them and to provide per allele coverage and per read set information. The coverage is then used to compute a phi score, i.e. a normalized chi-squared statistic varying between 0 and 1, which ranks best those SNPs that are discriminant between the samples. Finally, SNPs are sorted according to the phi score.
Time and memory consumption examples
Number of reads
Max. RAM use
H. sapiens tissues RNA-seq
S. cerevisiae WGS
S. cerevisiae WGS
S. cerevisiae WGS
Soil and seawater metagenomes
E. coli WGS
11 million and 0.08 million
S. cerevisiae WGS
2.25 million and 0.26 million
Variations in a transcriptome (including AS events) correspond to recognizable patterns in the dBG , known as ‘bubbles’ as briefly described in the DISCOSNP section above.
An example of such a bubble is given in Fig. 3.
The KISSPLICE program is composed of four steps: (i) de Bruijn graph construction, (ii) biconnected components decomposition, (iii) bubble enumeration, and (iv) event classification and quantification. In the first step, common to other Colib’read tools, the dBG is built from the set of reads using the GATB structure.
The second step in KISSPLICE decomposes the dBG into biconnected components (see  for formal definitions). This step requires marking the nodes and cannot be performed with the current version of the GATB structure, explaining why an explicit representation of the dBG is required. This decomposition has the advantage of not splitting the searched motifs while offering the possibility of performing the motif search in each component independently, and possibly in parallel. With RNA-seq data, this lossless graph decomposition is very efficient for splitting the dBG. For DNA-seq data, this decomposition is not efficient as most of the graph is made of a single biconnected component.
The third step, bubble enumeration, is the core of the KISSPLICE program. In this step the goal is to find all motifs (bubbles) satisfying the model constraints. This step is implemented using the enumeration algorithm given in .
Finally, in the fourth step each bubble is classified into four categories (indels, SNPs, AS events, and repeats) and quantified in each condition, independently. The quantification is done with KISSREADS, where we obtain the number of reads for each condition mapping to each variant. The final result, i.e. for each event the sequence of the variable part plus some sequence context and the quantification, is then output in a FASTA format.
TAKEABREAK  is a method of detecting inversion variants from one or several sets of reads without any reference genome. The rationale behind it is similar to that of DISCOSNP: inversion variants generate particular topological motifs in the dBG.
Inversion variants are defined as follows: a sequence I is said to be an inversion variant between two genomes if we can find the sequence aIb in one genome and aI’b in the other, with a and b being two k-mers and I’ being the reverse complement of I. We define the k-mers u and v as the first and last k-mers of I respectively. The occurrence in the data of the four breakpoint sequences au, vb, av’, and u’b (each of size 2k) generate a particular motif in the dBG that we call the inversion pattern. This motif is composed of two k-forks (a k-fork can be defined by two paths of size k joined at one extremity by a common branching node) joining together in a pseudo-cycle the four k-mers a, u, v, and b. An efficient algorithm was implemented in the software TAKEABREAK to find such inversion breakpoints in the dBG while avoiding numerous false positives due to repeated sequences. The implementation has very limited memory impact and runtimes (Illumina reads simulated at 2×40x coverage from human chromosome 22 can be treated in less than 10 min, with less than 1 GB of memory).
MAPSEMBLER2  is a targeted assembly program. It takes as input one or more set(s) of NGS raw reads (FASTA or FASTQ, gzipped or not) and a set of input sequences, called the ‘starters’. All the input read sets are used together to assemble the neighbors of each of the starters provided. These neighbors are output either as simple sequences (a sequence is cut as soon as two choices occur during the assembly) or as a graph in which polymorphisms are shown.
The visualization framework was designed to facilitate the interpretation of MAPSEMBLER2 outputs. Currently, this visualizer is compatible only with JSON format MAPSEMBLER2 outputs and with any dBG graph respecting the specific JSON characteristics. Further work will make GSV compatible with semantic web or systems biology tools to visualize, for example, RDF files or biological networks.
COMMET  provides a global similarity overview between all datasets of a large metagenomic project. Directly from non-assembled reads, all-against-all comparisons are performed through an efficient indexing strategy. The results are stored as bit vectors, i.e. a compressed representation of read files, which can be used to further combine read subsets by common logical operations. Finally, COMMET computes a clusterization of the metagenomic datasets, which is visualized through dendrograms and heatmaps.
LORDEC  is a tool to correct sequencing errors in long reads obtained from third-generation high-throughput sequencing technologies . Third-generation sequencing machines, especially PacBio, offer the advantage of delivering much longer reads than previous technologies (up to 20 Kb), often at the expense of sequence precision. Current estimates show that sequencing errors in PacBio reads average around 15 %, while traditional Illumina short reads exhibit an average error rate around 1 %. Moreover, PacBio sequencing suffers from a majority of insertion/deletion errors. It is thus necessary to correct these long reads before analysis, or at least during assembly, and different solutions have been proposed [19–22], but these approaches “require high computational resources and long running times on a supercomputer even for bacterial genome datasets”. .
In summary, LORDEC adopts a hybrid approach that takes advantage of the low error rate of short reads to correct the long ones. To avoid the computational bottleneck of all-against-all alignments, LORDEC builds the dBG of the short reads, and aligns long reads to the paths of the dBG. LORDEC exploits the fact that the dBG summarizes in a single structure the layout of short reads along the target DNA/RNA sequence. Hence, aligning a long read to the dBG allows the correction of erroneous sequence positions much more efficiently, and in a scalable manner. The dBG implementation of GATB allows LORDEC to process huge short reads libraries and to scale up to vertebrate or plant genome cases.
The LORDEC software offers several programs: the main one for correcting the long reads, and others for trimming and splitting corrected reads into corrected regions if needed. The output distinguishes these with lower vs upper cases. The value of k can be optimized by trying different values as low as possible around log4 (genome-length) (see  for an explanation).
Note that unlike the other tools, LORDEC addresses sequencing errors (which can be seen as technical artifacts) rather than biological events (such as variants, AS, etc.). Hence, LORDEC fulfills a need for preprocessing the read data before further analyses can extract biological information from it, as illustrated by the application to genome assembly described below.
Table 2 presents several time and memory footprint results, showing how the tools presented scale up on large raw datasets.
De novo identification of alternative splicing events in human RNA-seq data with KISSPLICE
KISSPLICE was applied to a human dataset that consists of 32 million reads from human brain and 39 million reads from liver from the Illumina Body Map 2.0 Project (downloaded from the Sequence Read Archive, study accession number ERP000546, brain read accession numbers ERR030882 and ERR030890, liver read accession numbers ERR030887 and ERR030895). Even though KISSPLICE does not require a reference genome, we applied it to a case where an annotated reference genome is indeed available in order to be able to assess whether our predictions are correct.
KISSPLICE ran in three hours using less than 8 GB RAM and was able to identify 2,336 bubbles corresponding to AS events.
To assess whether these predictions were correct, we aligned both paths of each bubble to the human reference genome (version hg19) using STAR  with default parameters.
We found that for 132 bubbles (5.7 %), the two paths did not map to the same genomic location, suggesting that the bubbles were false positives. A manual inspection of these cases revealed that most of them were due to repeats.
Among the bubbles where both paths mapped to the same location, we found that 1,714 (81 %) corresponded to annotated AS events, according to Ensembl v75 annotation . In contrast, 398 (19 %) corresponded to putative novel AS events, with at least one splice site not annotated before. Out of those 398 cases, 78 % (vs 97 % of them for the annotated splice sites) were canonical (GT-AG), and 22 % were non-canonical. An issue common to all transcriptome assemblers is that genomic indels, when located in transcribed regions, can be confused with AS events since they also generate bubbles in the dBG. In the presence of a reference genome, we can tell them apart easily as one path will map in two blocks and the other in one block. Using this criterion, we found that half of the non-canonical novel AS events were indeed indels (49 %). The remaining 51 % were composed of GC-AG, novel splice sites, and some mapping or assembly errors.
To summarize, we find that out of 2,336 bubbles reported by KISSPLICE, 76 % are AS events annotated in Ensembl, 14 % are AS events, not annotated but canonical, 2 % are AS events, not annotated and not canonical, 2 % are genomic indels, and 5 % are repeat-associated false positives.
Furthermore, we reported in  that KISSPLICE is more sensitive than Trinity, a widely used full-length transcriptome assembler, for calling AS events. The recent developments  in KISSPLICE show a particular sensitivity enhancement compared to Trinity in the case where there is still some pre-mRNA in the sample to be analyzed. The presence of pre-mRNA can in practice vary from 5 – 15 % depending on the protocol to isolate mRNA (total RNA vs nuclear, polyA+ vs polyA-). This can have a large impact on the assembly since introns are usually repeat-rich and the repeats are currently poorly handled by transcriptome assemblers.
Assessing DISCOSNP recall on real read sets from Saccharomyces cerevisiae
Isolated SNPs found in S. cerevisiae and validated in 
First population studied
(5 found among 6)
Predicted by DISCOSNP
Second population studied
(9 found among 9)
Predicted by DISCOSNP
Third population studied
(13 found among 14)
Predicted by DISCOSNP
Targeted assembly of S. cerevisiae using MAPSEMBLER2
Metagenomics global similarity overview of five gut metagenomes with COMMET
The MetaSoil study focuses on untreated soils of the Park Grass Experiment, Rothamsted Research, Hertfordshire, UK. One of the goals of this study is to assess the influence of depth, seasons, and extraction procedure on the sequencing . To achieve this, the 13 metagenomes from MetaSoil, two additional soil metagenomes, and a seawater metagenome were compared at the functional level using MG-RAST . This approach identified 835 functional subsystems present in at least one of the metagenomes that were used for clustering samples.
This study was reproduced with COMMET on all available metagenomes. The generated bit vectors total 68 MB, while the explicit representation of the FASTA results requires 6.4 GB. The storage footprint is thus divided by a factor of 100. This ratio is even higher if using FASTQ format or if dealing with larger read files. The COMMET computation time was 828 min.
LORDEC: impact of read correction on genome assembly
To summarize, experiments on real data taken from bacterial and yeast species, up to the case of a vertebrate genome, show that LORDEC achieves a quality at least as high as that of available state-of-the-art methods, while usually being an order of magnitude faster. Exploiting the GATB implementation of the dBG makes it by far the most scalable and economical option in terms of memory usage: LORDEC can process large datasets on a standard computer.
Datasets used to evaluate the efficiency and impact of LoRDEC read correction on the assembly
K-12 substr. MG1655
Number of reads
Average read length
Max. read length
Number of bases
Number of reads (millions)
Number of bases
Comparison of the assemblies obtained for E. coli and S. cerevisiae from either uncorrected or corrected PacBio reads
E. coli (k=64)
S. cerevisiae (k=51)
Number of contigs
Number of contigs ≥ 1 kbp
Genome coverage (%)
Total length (Mbp)
Largest contig (bp)
With uncorrected reads, and whatever the value of k, ABySS (v1.3.2) yields an assembly whose N50 value is close to k on both genomes, and with the longest contig below 400 bp. With PacBio reads corrected using LoRDEC, ABySS assemblies cover respectively 98 and 91 % of E. coli and S. cerevisiae genomes with respectively 321 and 1,657 contigs larger than 1 kbp (k=64 and k=51). Their N50 values reach 23 and 6.9 kbp respectively, while the largest contigs are 93 and 52 kbp long. Moreover, when aligning (using BLASTN, NCBI-BLAST-2.2.29+, with a reward of 1 and a penalty of −3) the contigs longer than 1 kbp against the reference genome, only 2.6 % of yeast contigs lack similarity, while all contigs of E. coli could be aligned.
The genome coverage and N50 values show that ABySS did not succeed in assembling any of the uncorrected PacBio datasets, while it yields satisfactory assemblies (i.e. sets of contigs) with the same PacBio reads that were corrected with LoRDEC . Without correction, an assembly obtained from a traditional program is useless; clearly, LoRDEC correction has a strong impact on the assembly quality of PacBio reads. Moreover, the correction is faster than the assembly and simplifies the latter. Importantly, the results suggest that hybrid correction using LoRDEC makes PacBio reads amenable to a classical de Bruijn graph assembly approach. LoRDEC has also been applied to MinION reads obtained with Oxford Nanopore technology and we observed a strong improvement in the mapability of the reads once corrected. More precisely, mapping the reads of an E. coli dataset on the reference genome with NucMer/Quast , we found that, while none of the raw MinION reads could be fully aligned, 2383 out of 2749 corrected reads were fully aligned on the genome, thereby covering 96 % of the genome. In addition, corrected reads could also be assembled with a de Bruijn-based approach.
Tool suite sharing
GUGGO Tool Shed [12, 13] was used to disseminate Colib’read Galaxy repositories. Corresponding tools are installed on our production Galaxy instance [35, 36], allowing scientists to use Colib’read tools freely after registration on the GenOuest core facility . As we join a dependencies package to our tools, Galaxy instance administrators can easily install either Galaxy tools (i.e. description files and wrappers) or Colib’read binaries and dependencies without any command line typing.
We propose bioinformatics tools dedicated to raw NGS data analyses for DNA-seq, RNA-seq and metagenomics studies. Thanks to the Galaxy platform, we easily made this tools suite available to life scientists, regardless their level of programming skills. Colib’read tools thus inherit reproducibility and accessibility support from Galaxy. Moreover, with the growing number of bioinformatics core facilities hosting Galaxy servers, tool shed usage enhances tools descriptors, binaries and dependencies sharing. This tools suite allows life scientists to find candidates that cannot be found with classical assembly-based approaches. Moreover, the algorithm developments described in this paper enhance the optimization and the management of the use of computing resources, in a time where these resources can not match the pace imposed by the NGS data deluge. Applications presented in this paper illustrate the low memory footprint of the six tools developed within the Colib’read framework, as well as their scalability. In replacement or combination with classical approaches, we thus propose to deal with higher amounts of information by using efficient computation strategies for NGS data.
Project name: Colib’read project Project home page:  Operating system(s): Platform independent Programming language: C++ Other requirements: GATB core License: A-GPL and CeCILL Any restrictions to use by non-academics: None
All data sets supporting the analyses are available from the GigaScience GigaDB repository .
This work is supported by ANR Colib’read; ANR-12-BS02-0008). ER, AZEA, BC acknowledge the support of Défi MASTODONS SePhHaDe from CNRS, Labex NUMev, and Projet Investissements d’Avenir France Génomique. LS is supported by the Academy of Finland (grant number 267591). GS was partially supported by the ERC program FP7/2007-2013 / ERC grant agreement no. 10. YLB and OC were supported by the Brittany and Pays de la Loire regions through the Biogenouest network. We also acknowledge the use of the GenOuest bioinformatics platform at the University of Rennes 1.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Colib’read Web Site. http://colibread.inria.fr/. Accessed date 23 Nov 2015.
- Sacomoto G, Kielbassa J, Chikhi R, Uricaru R, Antoniou P, Sagot M, et al. KISSPLICE: de-novo calling alternative splicing events from rna-seq data. BMC Bioinforma. 2012; 13(S-6):5.Google Scholar
- Peterlongo P, Chikhi R. BMC Bioinforma. 2012; 13(1):48. doi:http://dx.doi.org/10.1186/1471-2105-13-48.
- Uricaru R, Rizk G, Lacroix V, Quillery E, Plantard O, Chikhi R, et al.Reference-free detection of isolated snps. Nucleic Acids Res. 2014. doi:http://dx.doi.org/10.1093/nar/gku1187, http://nar.oxfordjournals.org/content/early/2014/11/16/nar.gku1187.full.pdf+html. Accessed date 23 Nov 2015.
- Lemaitre C, Ciortuz L, Peterlongo P. Mapping-free and assembly-free discovery of inversion breakpoints from raw NGS reads In: Dediu A-H, Martín-Vide C, Truthe B, editors. Algorithms for Computational Biology. Tarragona, Spain: 2014. p. 119–30. doi:http://dx.doi.org/10.1007/978-3-319-07953-0_10. https://hal.inria.fr/hal-01063157. Accessed date 23 Nov 2015.
- Maillet N, Collet G, Vannier T, Lavenier D, Peterlongo P. COMMET : comparing and combining multiple metagenomic datasets. In: BIBM: 2014.Google Scholar
- Salmela L, Rivals E. Lordec: accurate and efficient long read error correction. Bioinformatics. 2014. doi:http://dx.doi.org/10.1093/bioinformatics/btu538 http://bioinformatics.oxfordjournals.org/content/early/2014/08/27/bioinformatics.btu538.full.pdf+html.
- Galaxy Web Site. http://usegalaxy.org. Accessed date 23 Nov 2015.
- Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, Shah P, et al. Galaxy: a platform for interactive large-scale genome analysis. Genome Res. 2005; 15(10):1451–1455.View ArticlePubMedPubMed CentralGoogle Scholar
- Goecks J, Nekrutenko A, Taylor J, Team TG. Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol. 2010; 11(8):86.View ArticleGoogle Scholar
- Blankenberg D, Kuster GV, Coraor N, Ananda G, Lazarus R, Mangan M, et al. Galaxy: A web-based genome analysis tool for experimentalists. Curr Protoc Mol Biol. 2010:19–10.Google Scholar
- GUGGO Web Site. https://www.e-biogenouest.org/groups/guggo. Accessed date 23 Nov 2015.
- GUGGO Galaxy Tool Shed. http://toolshed.genouest.org/. Accessed date 23 Nov 2015.
- Drezen E, Rizk G, Chikhi R, Deltel C, Lemaitre C, Peterlongo P, et al. Gatb: Genome assembly & analysis tool box. Bioinformatics. 2014; 30(20):2959–961.View ArticlePubMedPubMed CentralGoogle Scholar
- Chikhi R, Rizk G. Space-efficient and exact de bruijn graph representation based on a bloom filter. Algorithm Mol Biol. 2013; 8:22.View ArticleGoogle Scholar
- Salikhov K, Sacomoto G, Kucherov G. Using cascading bloom filters to improve the memory usage for de brujin graphs. Algorithm Mol Biol. 2014; 9:2.View ArticleGoogle Scholar
- Grabherr MG, Haas BJ, Yassour M, Levin JZ, Thompson DA, Amit I, et al. Full-length transcriptome assembly from RNA-Seq data without a reference genome. Nat Biotechnol. 2011; 29:644–52. doi:http://dx.doi.org/10.1038/nbt.1883.View ArticlePubMedPubMed CentralGoogle Scholar
- Sacomoto G, Sinaimeri B, Marchet C, Miele V, Sagot M, Lacroix V. Navigating in a sea of repeats in rna-seq without drowning. In: 14th International Workshop on Algorithms in Bioinformatics (WABI): 2014. p. 82–96.Google Scholar
- Koren S, Schatz MC, Walenz BP, Martin J, Howard JT, Ganapathy G, et al. Hybrid error correction and de novo assembly of single-molecule sequencing reads. Nat Biotechnol. 2012; 30(7):693–700.View ArticlePubMedPubMed CentralGoogle Scholar
- Bashir A, Klammer AA, Robins WP, Chin CS, Webster D, Paxinos E, et al. A hybrid approach for the automated finishing of bacterial genomes. Nat Biotechnol. 2012; 30(7):701–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Au KF, Underwood JG, Lee L, Wong WH. Improving pacbio long read accuracy by short read alignment. PLoS ONE. 2012; 7(10):46679.View ArticleGoogle Scholar
- Deshpande V, Fung EDK, Pham S, Bafna V. Cerulean: A hybrid assembly using high throughput short and long reads. In: WABI. Springer: 2013. p. 349–63. LNCS.Google Scholar
- Philippe N, Boureux A, Tarhio J, Bréhélin L, Commes T, Rivals E. Using reads to annotate the genome: influence of length, background distribution, and sequence errors on prediction capacity. Nucleic Acids Res. 2009; 37(15):104.View ArticleGoogle Scholar
- Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. Star: ultrafast universal rna-seq aligner. Bioinformatics. 2013; 29(1):15–21. doi:http://dx.doi.org/10.1093/bioinformatics/bts635.View ArticlePubMedGoogle Scholar
- Flicek P, Amode MR, Barrell D, Beal K, Billis K, Brent S, et al. Ensembl 2014. Nucleic Acids Res. 2014; 42(D1):749–55. doi:http://dx.doi.org/10.1093/nar/gkt1196. http://nar.oxfordjournals.org/content/42/D1/D749.full.pdf+html.
- Kvitek DJ, Sherlock G. Whole genome, whole population sequencing reveals that loss of signaling networks is the major adaptive strategy in a constant environment. PLoS Genet. 2013; 9(11):1003972. doi:http://dx.doi.org/10.1371/journal.pgen.1003972.View ArticleGoogle Scholar
- Delmont TO, Robe P, Cecillon S, Clark IM, Constancias F, Simonet P, et al. Accessing the soil metagenome for studies of microbial diversity. Appl Environ Microbiol. 2011; 77(4):1315–1324.View ArticlePubMedGoogle Scholar
- Meyer F, Paarmann D, D’Souza M, Olson R, Glass E, Kubal M, et al. The metagenomics RAST server – a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinforma. 2008; 9(1):386. doi:http://dx.doi.org/10.1186/1471-2105-9-386.View ArticleGoogle Scholar
- Delmont TO, Prestat E, Keegan KP, Faubladier M, Robe P, Clark IM, et al. Structure, fluctuation and magnitude of a natural grassland soil metagenome. ISME J. 2012; 6(9):1677–1687. doi:http://dx.doi.org/10.1038/ismej.2011.197.View ArticlePubMedPubMed CentralGoogle Scholar
- Simpson JT, Wong K, Jackman SD, Schein JE, Jones SJM, Birol I. Abyss: A parallel assembler for short read sequence data. Genome Res. 2009; 19(6):1117–1123.View ArticlePubMedPubMed CentralGoogle Scholar
- Gurevich A, Saveliev V, Vyahhi N, Tesler G. Quast: quality assessment tool for genome assemblies. Bioinformatics. 2013; 29(8):1072–1075. doi:http://dx.doi.org/10.1093/bioinformatics/btt086, http://bioinformatics.oxfordjournals.org/content/29/8/1072.full.pdf+html.
- Galaxy Wiki. https://wiki.galaxyproject.org. Accessed date 23 Nov 2015.
- Galaxy Tool Shed. https://wiki.galaxyproject.org/ToolShed. Accessed date 23 Nov 2015.
- Andrieux A, Peterlongo P, Le Bras Y, Monjeaud C, Deltel C. Integrating new visualization tool in galaxy. In: Galaxy Community Conference 2014 (GCC2014): 2014.Google Scholar
- GenOuest Galaxy Platform. http://galaxy.genouest.org/. Accessed date 23 Nov 2015.
- Le Bras Y, Roult A, Monjeaud C, Bahin M, Quenez O, Heriveau C, et al. Towards a life sciences virtual research environment: An e-science initiative in western france. In: JOBIM 2013 Proceedings: 2013. p. 97–106. https://www.e-biogenouest.org/resources/128.
- GenOuest Web Site. http://www.genouest.org/. Accessed date 23 Nov 2015.
- Colib’read on Galaxy Web Site. http://colibread.inria.fr/colibread-on-galaxy/. Accessed date 23 Nov 2015.
- Bras YL, Collin O, Monjeaud C, Lacroix V, Rivals E, Lemaitre C, et al. Software and supporting data for Colib’read on Galaxy. GigaScience Database. 2016. doi:http://dx.doi.org/10.5524/100170.
- Yeast Genome Download Link. http://downloads.yeastgenome.org/sequence/S288C_reference/genome_releases. Accessed date 23 Nov 2015.