To assess the potential of protein function prediction in environmental genomics data, we analyzed shotgun sequences from four diverse and complex habitats. Using homology searches as well as customized gene neighborhood methods that incorporate intergenic and evolutionary distances, we inferred specific functions for 76% of the 1.4 million predicted ORFs in these samples (83% when nonspecific functions are considered). Surprisingly, these fractions are only slightly smaller than the corresponding ones in completely sequenced genomes (83% and 86%, respectively, by using the same methodology) and considerably higher than previously thought. For as many as 75,448 ORFs (5% of the total), only neighborhood methods can assign functions, illustrated here by a previously undescribed gene associated with the well characterized heme biosynthesis operon and a potential transcription factor that might regulate a coupling between fatty acid biosynthesis and degradation. Our results further suggest that, although functions can be inferred for most proteins on earth, many functions remain to be discovered in numerous small, rare protein families.

The analysis of completely sequenced genomes uncovers an astonishing variability between species in terms of gene content and order. During genome history, the genes are frequently rear-ranged, duplicated, lost, or transferred horizontally between genomes. These events appear to be stochastic, yet they are under selective constraints resulting from the functional interactions between genes. These genomic constraints form the basis for a variety of techniques that employ systematic genome comparisons to predict functional associations among genes. The most powerful techniques to date are based on conserved gene neighborhood, gene fusion events, and common phylogenetic distributions of gene families. Here we show that these techniques, if integrated quantitatively and applied to a sufficiently large number of genomes, have reached a resolution which allows the characterization of function at a higher level than that of the individual gene: global modularity becomes detectable in a functional protein network. In Escherichia coli, the predicted modules can be bench-marked by comparison to known metabolic pathways. We found as many as 74% of the known metabolic enzymes clustering together in modules, with an average pathway specificity of at least 84%. The modules extend beyond metabolism, and have led to hundreds of reliable functional predictions both at the protein and pathway level. The results indicate that modularity in protein networks is intrinsically encoded in present-day genomes.