Average genome size estimation improves comparative metagenomics and sheds light on the functional ecology of the human microbiome
Stephen Nayfach, Katherine S Pollard
Read Average genome size estimation improves comparative metagenomics and sheds light on the functional ecology of the human microbiome by Stephen Nayfach, Katherine S Pollard in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4389708.txt.
Introduction
Stephen Nayfach and Katherine Pollard built a tool that reads the average genome size of a microbial community, in Genome Biology. In shotgun metagenomics, DNA from a whole community is read at once, and the typical genome size of its members affects every comparison between samples. Their program, MicrobeCensus, estimates that average quickly and accurately from the read data. They applied it to 1,352 human microbiome samples from across the body.
Average genome size differed clearly between body sites and even within one site, and it tracked major differences in species mix and function. In the gut, larger average genomes went with more Bacteroides and more carbohydrate-use genes. The team also showed that ignoring genome size biases gene comparisons between samples, and that correcting for it improves detection of truly differing genes.
Catalog record
| CREATOR | Stephen Nayfach, Katherine S Pollard |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 83 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Average genome size estimation improves comparative metagenomics and sheds light on the functional ecology of the human microbiome By Stephen Nayfach, Katherine S Pollard Published in: Genome Biology DOI: 10.1186/s13059-015-0611-7 ABSTRACT Abstract Average genome size is an important, yet often overlooked, property of microbial communities. We developed MicrobeCensus to rapidly and accurately estimate average genome size from shotgun metagenomic data and applied our tool to 1,352 human microbiome samples. We found that average genome size differs significantly within and between body sites and tracks with major functional and taxonomic differences. In the gut, average genome size is positively correlated with the abundance of Bacteroides and genes related to carbohydrate metabolism. Importantly, we found that average genome size variation can bias comparative analyses, and that normalization improves detection of differentially abundant genes. Electronic supplementary material The online version of this article (doi:10.1186/s13059-015-0611-7) contains supplementary material, which is available to authorized users. BACKGROUND Shotgun metagenomics is increasingly being used to characterize the functional composition of microbial communities from the human body and many environments [1-4]. A common goal of these studies is to quantify gene family abundance and identify microbial genes or pathways that differ in abundance between environments, host phenotypes, or experimental conditions. Comparative metagenomics has shed light on how microbes have functionally adapted to the myriad of environments on Earth [4], and how variation in the functional composition of microbial communities can impact human health and disease [5,6]. Functional variation can be interpreted in the context of taxonomic variation estimated from the same shotgun data, shedding light on evolutionary and ecological processes. These studies require accurate quantification of gene abundance from microbial community DNA. While often overlooked, it is important to consider the average genome size (AGS) of cells in a microbial community when performing comparative metagenomic analyses. From a statistical perspective, AGS can be a potential source of bias when comparing the abundance of genes between communities. Specifically, the probability of sampling a gene from a community is inversely proportional to the AGS of that community [7]. If not accounted for, differences in AGS between samples can lead to the appearance of variation among genes that are present at equal copy number per cell (that is, false positives) and the appearance of stability among genes that vary in copy number per cell (that is, false negatives). Along with library size (that is, sequencing depth) and gene length, AGS should be accounted for in comparative metagenomic studies to identify ecologically meaningful genomic differences between microbial communities. AGS is also important for understanding the ecological and evolutionary forces acting on microorganisms within an environment. From an ecological perspective, microbial genome size may reflect environmental complexity, metabolic lifestyle, and community niche [8-12]. For example, in the gut it is believed that organisms with larger genomes follow more generalist lifestyles whereas those with smaller genomes are more specialized [8]. To illustrate this, Bacteroides thetaiotamicron (6.5 Mb) has a large genome with the metabolic potential to utilize a variety of both diet and host-derived glycans, whereas Methanobrevibacter smithii (1.9 Mb) is a methanogen specialized for utilization of H2. From an evolutionary perspective, reduced genome size may reflect genetic drift in small populations or genome streamlining in large populations [13,14]. Genetic drift can shape the size of microbial genomes by the fixation of mutations, which are biased towards deletions in Bacteria [15]. In contrast, genome streamlining can promote reduced genome size by selecting for the efficient use of nutrients in large microbial populations where nutrients limit growth [14]. Up to this point it has not been possible to rapidly and accurately estimate the AGS of microbial communities due to a lack of software designed for modern metagenomics data. This has limited our understanding of both the extent and impact of AGS variation in many environments, including the human microbiome. The only publicly available software tool, GAAS [10], estimates AGS based on BLAST searches of shotgun sequences against a database of microbial genomes. Given the large and increasing scale of both metagenomic data and reference genomes, and the relatively slow speed of BLAST, this method is not computationally practical (Results). Additionally, microbial communities frequently contain high fractions of 'novel' organisms, which have not been cultured or sequenced. Even in the well-studied human gut microbiome, it has been estimated that, on average, 43% of species abundance and 58% of richness cannot be captured by current microbial reference genomes [16]. It is not clear whether GAAS is able to accurately estimate AGS for metagenomes composed of novel taxa. Raes et al. [9] proposed to address these issues by estimating average genome size based on the density of reads assigned to a set of 35 essential single-copy genes using BLASTX. While significantly faster and less dependent upon community composition, this method was not designed for reads shorter than 300 base pairs (bp) and no software was released. Although current sequencing technologies are beginning to generate longer reads, there remains a huge volume of existing short-read data; for example, the MetaHIT [1,17-19] and Human Microbiome Project (HMP) [20] projects have together generated over 130 billion metagenomic reads from the human microbiome with an average length of only 96 bp [21]. It is not clear whether the Raes method is able to estimate AGS for these modern short-read libraries. Others [22,23] have described similar methods, but these have not been extensively validated or made available as software. To address this problem, we developed MicrobeCensus to rapidly and accurately estimate AGS from shotgun metagenomic data and applied our tool to 1,352 human microbiome samples. We adopted a similar approach to Raes et al., but made significant methodological improvements that allowed us to accurately estimate AGS using reads as short as 50 bp. We found that AGS differs significantly both within and between body sites in the human microbiome and tracks with major functional and taxonomic differences between communities. For example, the AGS of stool metagenomes ranged from 2.5 to 5.8 megabases (Mb) and was positively correlated with the abundance of Bacteroides and genes related to metabolism, biosynthesis, and two-component systems, whereas Firmicutes and genes related to membrane transport were more abundant in metagenomes with smaller AGS. Furthermore, we confirmed that AGS is a major bias in comparative metagenomic analyses, and that normalization improves detection of differentially abundant genes. These discoveries could not have been made without the development of MicrobeCensus. RESULTS AND DISCUSSION Estimation of average genome size from metagenomic data At the core of our method is a database of 30 essential single-copy gene families found in nearly all Bacteria and Archaea (Additional file 1). These are a subset of the 40 bacterial and archaeal PhyEco markers identified by Wu, Jospin and Eisen [24], and they largely overlap with gene sets previously identified by Ciccarelli et al. [25], Raes et al. [9], and Wu et al. [26]. A preliminary analysis shows that most of these gene families are also present in Fungi, although are not as universal or stable as in Bacteria and Archaea (Additional files 1 and 2). Hence, we can expect that the vast majority of cells in a microbial community - particularly those dominated by Bacteria or Archaea - will carry one copy of each of these genes.