Section: New Results

HTS data processing

Providing end-user solutions, example from the Colib'read on galaxy project

Participants : Claire Lemaitre, Camille Marchet, Pierre Peterlongo.

Colib’read tools suite uses optimized reference-free algorithms for various analyses of NGS datasets, such as variant calling or read set comparisons. To facilitate data analysis and tools dissemination, we developed Galaxy tools and tool shed repositories. The galaxy package, facilitates the analyzis of raw NGS data for a broad range of life scientists [16].

Assembly of Streptococcus Bacteria

Participant : Dominique Lavenier.

With the microbiological and bacteriological group of the Rennes hospital, we design a new strategy to assemble the genomes of 40 Streptococcus bacteria. Each strain has been sequenced and independently assembled using different assembly tools. For a specific strain, a merge of the contigs is done using the MIX software. This step allows the number of contigs to be significantly reduced, resulting in a better final assembly compared to each individual assembly. The comparison with other known Streptococcus genomes indicates where phages are located in the genome [20].

Data-mining applied to GWAS

Participants : Pham Hoang Sun, Dominique Lavenier.

Identifying variant combination association with disease is a bioinformatics challenge. This problem can be solved by discriminative pattern mining that uses statistical functions to evaluate the significance of individual biological patterns. There is a wide range of such measures. However, selecting an appropriate measure as well as a suitable threshold in some specific practical situations is a difficult task. We propose to use the skypattern technique which allows combinations of measures to be used to evaluate the importance of variant combinations without having to select a given measure and a fixed threshold. Experiments on several real variant datasets demonstrate that the skypattern method effectively identifies the risk variant combinations related to diseases  [28].

Variant detection in transcriptomic data

Participant : Camille Marchet.

We defined a method to identify, quantify and annotate SNPs (Single Nucleotide Polymorphisms) using RNA-seq reads only. Organisms with a poor quality or no reference genome can take benefit of this approach, as well as studies where not enough material is available for sequencing from one individual, where samples can be pooled. The method relies on motifs discovery and post-treatment in de Bruijn graphs built from the reads. It can be used for any species to annotate SNPs and predict their impact on proteins as well as test their association to a phenotype of interest. The approach has been validated using well known human RNA-seq data. Results have been compared with state of the art approaches for variant calling. We showed that the methods perform similarly in terms of precision and recall. Then we focused on the main target of the study, namely the non-model species. We finally validated experimentally the predictions of our method [18].

Faster de Bruijn graph compaction

Participant : Antoine Limasset.

We developed a new algorithm, called BCALM2, for the compaction of de Bruijn graphs. BCALM2 is a parallel algorithm based on minimizer repartition of sequences. This repartition allows the compaction of extremely large graphs with moderate memory usage and time. The compaction of a human sequencing graph can be done in 1 hour with only 3GB of memory and huge genomes, such as the pine and white spruce ones (more than 20Gbp each), can be handled using our approach on a regular server (2 days and 40GB of memory). Those results argue that BCALM2 is one order of magnitude more efficient than available approaches and can tackle the assembly bottleneck of constructing a compacted de Bruijn graph [14].

Scaffolding

Participants : Rumen Andonov, Sebastien François, Dominique Lavenier.

We developed a method for solving genome scaffolding as a problem of finding a long simple path in a graph defined by the contigs that satisfies additional constraints encoding the insert-size information. Then we solved the resulting mixed integer linear program to optimality using the Gurobi solver. We tested our algorithm on several chloroplast genomes and showed that it outperforms other widely-used assembly solvers by the accuracy of the results [25].