Abstract
BACKGROUND: Advances in sequencing, assembly, and assortment of contigs into species-specific bins has enabled the reconstruction of genomes from metagenomic data (MAGs). Though a powerful technique, it is difficult to determine whether assembly and binning techniques are accurate when applied to environmental metagenomes due to a lack of complete reference genome sequences against which to check the resulting MAGs. METHODS: We compared MAGs derived from an enrichment culture containing ~20 organisms to complete genome sequences of 10 organisms isolated from the enrichment culture. Factors commonly considered in binning software-nucleotide composition and sequence repetitiveness-were calculated for both the correctly binned and not-binned regions. This direct comparison revealed biases in sequence characteristics and gene content in the not-binned regions. Additionally, the composition of three public data sets representing MAGs reconstructed from the Tara Oceans metagenomic data was compared to a set of representative genomes available through NCBI RefSeq to verify that the biases identified were observable in more complex data sets and using three contemporary binning software packages. RESULTS: Repeat sequences were frequently not binned in the genome reconstruction processes, as were sequence regions with variant nucleotide composition. Genes encoded on the not-binned regions were strongly biased towards ribosomal RNAs, transfer RNAs, mobile element functions and genes of unknown function. Our results support genome reconstruction as a robust process and suggest that reconstructions determined to be >90% complete are likely to effectively represent organismal function; however, population-level genotypic heterogeneity in natural populations, such as uneven distribution of plasmids, can lead to incorrect inferences.