Common workflow language (CWL)-based software pipeline for de novo genome assembly from long- and short-read data

Background: Here, we created an automated pipeline for the de novo assembly of genomes from Pacific Biosciences long-read and Illumina short-read data using common workflow language (CWL). To evaluate the performance of this pipeline, we assembled the nuclear genomes of the eukaryotes Caenorhabditis elegans (similar to 100 Mb), Drosophila melanogaster (similar to 138 Mb), and Plasmodium falciparum (similar to 23 Mb) directly from publicly accessible nucleotide sequence datasets and assessed the quality of the assemblies against curated reference genomes. Findings: We showed a dependency of the accuracy of assembly on sequencing technology and GC content and repeatedly achieved assemblies that meet the high standards set by the National Human Genome Research Institute, being applicable to gene prediction and subsequent genomic analyses. Conclusions: This CWL pipeline overcomes current challenges of achieving repeatability and reproducibility of assembly results and offers a platform for the re-use of the workflow and the integration of diverse datasets. This workflow is publicly available via GitHub (https://github.com/vetscience/Assemblosis) and is currently applicable to the assembly of haploid and diploid genomes of eukaryotes.