A bioinformatic pipeline to elucidate the links between viruses and their hosts in microbial communities, applied to viruses in anaerobic digestion processes

International audience ; Viruses are key-players in microbial ecosystems. However, predicting hosts from viruses is still a major challenge in microbial ecology. A few in silico methods for metagenomics data have proven useful for this purpose (e.g. in [1]), and they are highly invaluable when studying environmental samples. We developed a bioinformatic pipeline including the detection of CRISPR protospacers in viral contigs, a method previously used to predict hosts from marine viruses [1]. We applied our pipeline to anaerobic digestion (AD) ecosystems, in the context of organic waste treatment and valorisation. We focused on the diversity of viruses infectingmethanogens, the latter being the key actors of methane production during AD. Viral diversity is only starting to be explored in AD processes [2], hence the great potential of new virus discovery in our study.After enrichment of methanogenic archaea in AD microcosms by growth on formate as the sole carbon source, 2 DNA metaviromes and 5 cellular metagenomes were sequenced using Illumina NextSeq. Our pipeline was applied to all the obtained data. It was executed on the cluster of the INRAE MIGALE bioinformatics platform. The most generic steps of our pipeline were scripted as a snakemake workflow, to favour reproducible and scalable data analysis (https://forgemia.inra.fr/cedric.midoux/workflow_metagenomics). After a preprocessing step, reads were assembled with metaSPADES. Coding regions were predicted with Prodigal. Taxonomic assignation of the contigs and of their predicted genes was obtained with kaiju against NCBI nr database. Functional annotation of the predicted genes was performed with Diamond against Phrogs(https://phrogs.lmge.uca.fr/), a database dedicated to prokaryotic viruses, and with ghostKoala against KEGG database (https://www.kegg.jp/ghostkoala/). For each dataset, reads were mapped to the assembled contigs.Several steps specifically dedicated to the prediction of hosts from viral contigs were performed using bash and python scripts. For ...