Metagenomic mining for microbiologists

Abstract

Microbial ecologists can now start digging into the accumulating mountains of metagenomic data to uncover the occurrence of functional genes and their correlations to microbial community members. Limitations and biases in DNA extraction and sequencing technologies impact sequence distributions, and therefore, have to be considered. However, when comparing metagenomes from widely differing environments, these fluctuations have a relatively minor role in microbial community discrimination. As a consequence, any functional gene or species distribution pattern can be compared among metagenomes originating from various environments and projects. In particular, global comparisons would help to define ecosystem specificities, such as involvement and response to climate change (for example, carbon and nitrogen cycle), human health risks (eg, presence of pathogen species, toxin genes and viruses) and biodegradation capacities. Although not all scientists have easy access to high-throughput sequencing technologies, they do have access to the sequences that have been deposited in databases, and therefore, can begin to intensively mine these metagenomic data to generate hypotheses that can be validated experimentally. Information about metabolic functions and microbial species compositions can already be compared among metagenomes from different ecosystems. These comparisons add to our understanding about microbial adaptation and the role of specific microbes in different ecosystems. Concurrent with the rapid growth of sequencing technologies, we have entered a new age of microbial ecology, which will enable researchers to experimentally confirm putative relationships between microbial functions and community structures.

Figure 1

(a) PCA based on the relative distribution of annotated sequences (E-value<10⁻⁵) categorized in 838 different functional subsystems detected in the 77 metagenomes. Distributions were normalized as a function of the number of annotated sequences for each metagenome. The percentages of the illustrated two major axes correspond to the fraction of the total variance that they represent (see insert showing all of the axes and their percentage of the overall variance). (b) Relationship between average sequence length and the percentage of annotated functions (E-value<10⁻⁵) for the metagenomes used here. The different average sequence sizes are due in part to variations in sequencing technology. In addition, ocean and Antarctic metagenomes have annotations varying considerably for the same average sequence length. This fluctuation is due in part to the presence of sequences related to eukaryotic and virus sequences for oceans and Antarctic aquatic environments.

Figure 2

Comparison of the relative distribution in percentage (based on the annotated sequences (E-value<10⁻⁵)) of five functional classes and one genus (SEED annotation) among the 77 metagenomes deposited in MG-RAST. The horizontal line corresponds to the average of the relative distribution for each of the 15 environments.

Figure 3

PCA of six selected ecosystems based on their number of sequences associated with petroleum hydrocarbon degradation functions (E-value<10⁻⁵). The functional classes as provided by MG-RAST and the local blasts are plotted on the same PCA as the samples in order to observe relationships between function and environment.