Speaker
Description
A bacteria cell contains several DNA molecules. One is the proper bacteria chromosome. Several others are short circular extrachromosomal DNA molecules called plasmids. Plasmids play a key role in antimicrobic resistance as bacteria can share them and because some plasmids can integrate the bacteria chromosome. Being able to differentiate the genomic content of the chromosome from the plasmids is an important task for further genomic, phylogenetic and epidemiologic analyses.
The sequencing of a bacteria sample provides fragments of the copies of the DNA molecules present in the bacteria cells (reads). The assembly process (like a puzzle solving) is producing an assembly graph where the vertices are contigs (continuous DNA fragments longer than the reads) and edges are potential connections between them the assembly process was not able to determine if they are true connections or not. The plasmid binning problem consists in grouping contigs into bins such that each bin is the set of all the contigs of only one plasmid. The main difficulty arises from the nature of the assembly: the plasmid binning is not a partitioning problem because one contig can belong to several plasmids, and sometimes can be shared between a plasmid and the chromosome. Such contigs are called repeats between the molecules. The assembly process returns contigs associated with a normalized coverage which is a proxy of the number of times the contig is present in the sample. Unfortunately, the analysis of the coverage distribution is not enough to solve the issues the repeats raise.
Here we address the plasmid binning problem as a combinatorial optimization problem. Knowing that plasmids are mainly circular molecules, and that their copies in the cell are more abundant than the chromosome molecule, we search for a set of circular flows that cover the best the contigs previously identified as potentially part of a plasmid by a classifier. As the number of plasmid molecules is unknown, we adopt a parsimonious approach to find the minimum number of flows that cover the best the plasmidic contigs. In addition, we provide solutions to bin the contigs even if the plasmid circularity is lost in the assembly graph due to missing edges. We model the plasmid binning problem with a mixed integer linear programming approach we have implemented as an installable and user-friendly Python3 package PlasBin-HMF. We tested our tool with the state of the art on 836 real bacteria samples and show very encouraging results. While we achieve the best results overall, we provide a differential analyses according to the species the samples are coming from to show the ability of other tools to achieve better results on some specific species.
| Pracovisko fakulty (katedra)/ Department of Faculty | Department of Computer Science |
|---|---|
| Tlač postru/ Print poster | Budem požadovať tlač /I hereby required to print the poster in faculty |