Meteor2: accurate taxonomic profiling of host-associated microbial communities for shotgun metagenomics

International audience

إنجليزي. Objective: Detection and quantification of microbial species is a fundamental task in metagenomics. Many taxonomic profilers are available for this purpose, each with their own strengths and weaknesses. Here, we introduce Meteor2, a new taxonomic profiler that provides accurate results, including subdominant species poorly accounted for by competing tools.Methods: Meteor2 aligns reads on non-redundant microbial gene catalogs covering a specific host-associated ecosystem. Currently, we provide catalogs for pig, mouse, chicken, rat, rabbit, dog, cat and human microbiota (gut, skin and oral cavity). In each catalog, co-abundant marker genes have been identified and are used to infer species abundance.To assess Meteor2 performance, we generated two simulated datasets mimicking respectively mouse (62 samples) or human fecal metagenomes (100 samples) using 5,114 high-quality genomes. Next, we computed taxonomic profiles of these samples with Meteor2 and seven other popular tools. Finally, we compared all tools with performance metrics assessing either presence/absence of species (sensibility, specificity, F1-score) or their abundance (Bray-Curtis dissimilarity).Results: Overall, Meteor2, sylph and MetaPhlan4 were the top-ranked tools. Notably, Meteor2 was the best at detecting and estimating abundance of subdominant species. We confirmed and illustrated these findings on real datasets.Conclusion: Meteor2 is a taxonomic profiler for shotgun metagenomics that provides more accurate results than state-of-the-art tools in many contexts. We are currently integrating new modules that provide functional and strain-level profiling. Meteor2 is freely available at https://github.com/metagenopolis/meteor.