Overview

The de novo assembly of raw sequence data is a key process when analysing data from shotgun metagenomic sequencing. It allows recovering draft genomes from a pool of mixed raw reads, yielding longer sequences that offer contextual genomic information and afford a more complete picture of the microbial community. It may also represent one of the greatest bottlenecks when obtaining trustworthy, reproducible results.

LMAS is an automated workflow enabling the benchmarking of traditional and metagenomic prokaryotic de novo assembly software using defined mock communities. The results are presented in an interactive HTML report where selected global and reference specific performance metrics can be explored.

LMAS requires as inputs the reference sequences (complete genomes) and short-read paired-end raw data. This raw data can be either obtained in silico by creating simulated reads from the reference complete genomes or from directly sequencing the references. Optionally, information can be passed, in markdown, on the input samples to be presented in the report. All complete genomes (reference linear replicons) should be provided in a single file.

The LMAS Workflow. The input sequencing data is assembled in parallel, resources permitting, by the set of assemblers in LMAS. The resulting assembled sequences are processed and assembly quality metrics are computed, both globally and in comparison to the reference sequences provided. The global and per reference metrics are grouped in the LMAS report for interactive exploration.

Implementation

LMAS was implemented in Nextflow to provide flexibility and ensure the transparency and reproducibility of the results. LMAS relies on the use of Docker containers for each assembler, allowing versions to be tracked easily.

Nextflow, a workflow management software, allows the effortless deployment of LMAS in any UNIX-based system, from local machines to high-performance computing clusters (HPCs) with a container engine installation, such as Docker, Shifter or Singularity.

The local installation of the LMAS workflow, including the Docker containers, requires 7.3 gigabytes (GB) of free disk space. The default requirements to execute the workflow are at least 32 GB of memory and 8 CPUs, with a maximum of 100 GB of memory and 32 CPU. This can be easily adjusted but might compromise the performance of the assemblers contained in LMAS. The assemblers can be skipped individually through the use of parameters. The disk space required for execution depends greatly on the size of the input data but, in average, LMAS generates approximately 17 GB of data per GB of input data.