Scalable and efficient data analysis pipelines for clinical metagenomics from nf-core

BACKGROUND Clinical metagenomics is an emerging field with the potential to transforminfectious disease diagnostics. Clinical metagenomics analysis often includesa series of initial steps, including rapid taxonomic identification ofhigh-throughput sequencing (HTS) reads against known reference databases, denovo genome assembly for identification of database-unrepresented taxa, andfunctional screening for characteristics such as antimicrobial resistance. As data throughput increases, clinical scientists need high-quality,efficient, automated pipelines that scale with increasing numbers of samplesand genomes, replicability to ensure data fidelity, and portability to handleexecution on a different computing infrastructure. METHODS The nf-core initiative aims to build community-based, software-developmentbest-practices and (bio)informatic pipelines written in the workflow managerlanguage Nextflow. A range of metagenomics-based pipelines have beendeveloped within the initiative to address primary analysis in a reproducibleand interoperable manner. Design decisions during the construction of thepipeline have drawn from established metagenomics procedures, but alsoexperiences from low-biomass ancient DNA metagenomics research. RESULTS I will introduce some of the nf-core pipeline offerings relevant for clinicalmetagenomics: (1) nf-core/taxprofiler for profiling microbial communitieswith potential for monitoring known and emerging pathogens, (2) nf-core/magfor metagenomic de novo genome assembly for identification of novel(potentially pathogenic) taxa, and (3) nf-core/funcscan for screening forgenes relating to antimicrobial resistance and natural product production. Writing pipelines in Nextflow provides efficiency through integration withjob schedulers on HPC or the cloud, automatic retry of failed jobs, andreproducibility through software containers. DISCUSSION I will describe how the various pipelines can be used in an interoperablemanner to produce an end-to-end set of results relevant for clinicians instandardised and unified outputs. I will also show how nf-core aims to be asuser-friendly as possible, through extensive documentation, commoncommand-line and graphical-user interfaces, and interactive and customisableresult summaries.