Tell TaxContigs: Microbial Metagenome Assembly, Taxonomy, and Abundance Estimation Using UST TELL seq Long range Information

Deconvolving diversity of a metagenome is critical for understanding the role of a given microbial community in human health and disease, small molecule biosynthesis, and other complex ecosystems where more reductive analysis proves to be elusive Metagenomic assembly is one common method for characterizing a metagenome, especially for identifying novel gene content or novel organisms However, analyzing sequencing data from microbial mixtures with a high dynamic range of relative abundance of strains, close relatedness, and repetitive genomic content amongst members can vastly complicate the genome assembly process of individual microorganisms and strains Furthermore, efficient assembly requires high fidelity sequencing reads to avoid ambiguities We previously developed a method that captures long range molecular origin information from kilobase long genomic fragments by a process of DNA barcoding that we called transposase enzyme linked long read sequencing (TELL seq) developed by Universal Sequencing Technologies ( 1 TELL seq barcoded fragments can be sequenced with instruments that process short reads (i e high fidelity sequencing) Here, we show that integration of TELL seq data with a computational pipeline that combines de novo genome assembly (Tell Link) with taxonomic classification and abundance estimation (Tell TaxContigs) provides highly accurate metagenomic analyses We show how the application of Tell Link and Tell TaxContigs on sequencing data generated from commercially available microbial mixture standards results in genome assemblies with contiguities larger than 1 Mbp (N 50 and highly accurate classification and relative abundance estimation for organisms at 0 18 or greater relative abundance, respectively Therefore, Tell Link, in combination with genome binning software e g metabat 2 2 provides highly contiguous and high fidelity genome assemblies of abundant organisms in a metagenomic sample Tell TaxContigs classifies contigs and unassembled reads with BLASTn and using a deep learning approach resolves ambiguities, rules out false positive classifications, and accurately estimates relative abundances of classified species This approach has an average margin of error of lower than 1 in enumerating relative abundance for the microbial mixture standards tested