Modeling association networks from metagenome time series - a proof-of-concept from two contrasting estuaries

Understanding microbial responses to environmental gradients is important for ecosystem assessment in estuarine environments. Here, we integrated time-series samples and shotgun metagenomics with association network inference, to examine patterns of taxonomic and functional co-variation. We collected and sequenced eDNA from two sediment time series from neighbouring but environmentally contrasting Basque estuaries, and compared the resulting taxonomic and functional association networks, based on Metagenome Assembled Genomes (MAGs) and KEGG Orthologous Groups (KOs), respectively. Networks inferred using 16S rRNA gene metabarcoding data were also included in this comparison. Several topology-based network properties differed strongly between the two sites, although differences were not always consistent when comparing the networks based on MAGs, KOs and metabarcoding. KO-based networks revealed distinct sets of associated Ortholog groups between sites, consistent with known environmental differences. Contrary to our expectations, metabolically related genes rarely formed complete pathway modules in gene networks, likely due to functional redundancy, gene copy number variation, synchronized pathway dynamics, or algorithmic constraints of correlation-based inference. Overall, the results illustrate both the potential and interpretative limitations of metagenome-based time association networks for functional ecosystem assessment, and support their use as hypothesis-generating tools that require further validation across a larger range of sites following environmental gradients.