| description |
Predicting the dynamics of community composition and functional attributes responding to environmental changes is an essential goal in community ecology but remains a major challenge, particularly in microbial ecology. Here, by targeting a low-diversity model system, we explored the spatial distribution of taxonomic and functional structure of 40 acid mine drainage (AMD) microbial communities across Southeast China profiled by 16S rRNA pyrosequencing and a comprehensive microarray (GeoChip). Similar environmentally dependent patterns of dominant microbial lineages and key functional genes were observed regardless of the large-scale geographical isolation. Functional and phylogenetic β-diversities were significantly correlated, while functional metabolic potentials were strongly influenced by environmental conditions and community taxonomic structure. Using advanced modeling approaches based on artificial neural networks, we successfully predicted the taxonomic and functional dynamics with significantly higher prediction accuracies of metabolic potentials (average Bray-Curtis similarity 87.8) as compared to relative microbial abundances (similarity 66.8), implying that natural microbial assemblages may be better predicted at the functional genes level rather than species. Furthermore, relative metabolic potentials of genes involved in many key ecological functions (e.g., nitrogen and phosphate utilization, metals resistance and stress response) were extrapolated to increase under more acidic and metal-rich conditions, indicating a critical strategy of stress adaptation in these extraordinary communities. Collectively, our findings indicate that natural selection plays a crucial role in shaping the taxonomic and functional patterns of AMD microbial community that readily predicted by modeling methods, and suggest that the inference-based approach is essential to better understand natural acidophilic microbial communities. |