Long read metagenomics

Long read (LR) DNA sequencing is revolutionizing comparative genomics of eukaryotes and microbial cultured isolates, but has not been widely adopted in microbial metagenomics due to limitations in throughput, cost, and data quality. Here, we developed a high-throughput, low cost molecular workflow and generated paired, long read DNA sequencing data from 47 complex, human fecal samples with the Pacbio (PB) Revio (591 Gbp), the Oxford Nanopore Technologies (ONT) PromethION (1171 Gbp), and 40 samples with Illumina (ILMN) synthetic long reads (625 Gbp). PB outperformed ONT in most sequencing metrics (% useable data, read quality, mean, median and read length N50), generated 2.43x more complete metagenome-assembled genomes (cMAGs) at the lowest cost per cMAG ($16 USD vs $25), and had the highest profiling classification rates resulting in detecting the most significant associations. Both PB and ONT generated 51 and 72x more cMAGs per Gbp than ILMN. We generated 985 (831 circular) cMAGs, performed whole genome and functional pangenome analyses across ten genera, and identified microbial genetic associations with various environmental and biological phenotypes related to undernutrition in toddlers from sub-Saharan Africa that would not have been possible with short reads.