@phdthesis{Hackl2016, author = {Hackl, Thomas}, title = {A draft genome for the Venus flytrap, Dionaea muscipula : Evaluation of assembly strategies for a complex Genome - Development of novel approaches and bioinformatics solutions}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-133149}, school = {Universit{\"a}t W{\"u}rzburg}, year = {2016}, abstract = {The Venus flytrap, \textit{Dionaea muscipula}, with its carnivorous life-style and its highly specialized snap-traps has fascinated biologist since the days of Charles Darwin. The goal of the \textit{D. muscipula} genome project is to gain comprehensive insights into the genomic landscape of this remarkable plant. The genome of the diploid Venus flytrap with an estimated size between 2.6 Gbp to 3.0 Gbp is comparatively large and comprises more than 70 \% of repetitive regions. Sequencing and assembly of genomes of this scale are even with state-of-the-art technology and software challenging. Initial sequencing and assembly of the genome was performed by the BGI (Beijing Genomics Institute) in 2011 resulting in a 3.7 Gbp draft assembly. I started my work with thorough assessment of the delivered assembly and data. My analysis showed that the BGI assembly is highly fragmented and at the same time artificially inflated due to overassembly of repetitive sequences. Furthermore, it only comprises about on third of the expected genes in full-length, rendering it inadequate for downstream analysis. In the following I sought to optimize the sequencing and assembly strategy to obtain an assembly of higher completeness and contiguity by improving data quality and assembly procedure and by developing tailored bioinformatics tools. Issues with technical biases and high levels of heterogeneity in the original data set were solved by sequencing additional short read libraries from high quality non-polymorphic DNA samples. To address contiguity and heterozygosity I examined numerous alternative assembly software packages and strategies and eventually identified ALLPATHS-LG as the most suited program for assembling the data at hand. Moreover, by utilizing digital normalization to reduce repetitive reads, I was able to substantially reduce computational demands while at the same time significantly increasing contiguity of the assembly. To improve repeat resolution and scaffolding, I started to explore the novel PacBio long read sequencing technology. Raw PacBio reads exhibit high error rates of 15 \% impeding their use for assembly. To overcome this issue, I developed the PacBio hybrid correction pipeline proovread (Hackl et al., 2014). proovread uses high coverage Illumina read data in an iterative mapping-based consensus procedure to identify and remove errors present in raw PacBio reads. In terms of sensitivity and accuracy, proovread outperforms existing software. In contrast to other correction programs, which are incapable of handling data sets of the size of D. muscipula project, proovread's flexible design allows for the efficient distribution of work load on high-performance computing clusters, thus enabling the correction of the Venus flytrap PacBio data set. Next to the assembly process itself, also the assessment of the large de novo draft assemblies, particularly with respect to coverage by available sequencing data, is difficult. While typical evaluation procedures rely on computationally extensive mapping approaches, I developed and implemented a set of tools that utilize k-mer coverage and derived values to efficiently compute coverage landscapes of large-scale assemblies and in addition allow for automated visualization of the of the obtained information in comprehensive plots. Using the developed tools to analyze preliminary assemblies and by combining my findings regarding optimizations of the assembly process, I was ultimately able to generate a high quality draft assembly for D. muscipula. I further refined the assembly by removal of redundant contigs resulting from separate assembly of heterozygous regions and additional scaffolding and gapclosing using corrected PacBio data. The final draft assembly comprises 86 × 10 3 scaffolds and has a total size of 1.45 Gbp. The difference to the estimated genomes size is well explained by collapsed repeats. At the same time, the assembly exhibits high fractions full-length gene models, corroborating the interpretation that the obtained draft assembly provides a complete and comprehensive reference for further exploration of the fascinating biology of the Venus flytrap.}, subject = {Venusfliegenfalle}, language = {en} }