Since Haeckel coined the term, ecology has advanced by studying mechanisms and feedbacks that govern how organisms interact with their environment and each other. From individuals to populations to communities to ecosystems, ecologists have designed reductionist experiments for strong inference and developed theory from first principles. Inference in microbiome science often runs in reverse, starting with gene surveys or meta-genomics of microbiomes, and making inferences from informatics, with little direct knowledge of the organisms. Many microbial sequences remain classified as biological “dark matter”, known only by the order of their nucleotides, their connection to organisms inferred from bioinformatics, and their roles in nature predicted from genes of organisms known in culture. While the ‘omics revolution holds great promise, it lags in connecting microbiome ‘omic descriptions with the ecological theories that predict organismal responses to environmental change. Bridging ‘omics insights and ecological theory has been hampered by three key problems: First, ‘omics techniques are only partially quantitative--insofar as they estimate relative abundances or proportions. Most ‘omic techniques were not designed to be capable of estimating process rates of a specific microbiome population, such as growth, mortality, and resource assimilation. Second, microbiomes are metabolically complex, using an array of electron donors and acceptors and associated reactions that far outnumber the basic biochemical reactions that support macrosystems. Third, living microbes are far more difficult to observe in their native environments than are macro-biota, challenging hypothesis-driven discovery of the species interactions that structure biological assemblages. Here, we present a new approach to detect and quantify microbial interactions, an approach that draws on ecological theory and concepts, with tools and techniques from isotope geochemistry, biochemistry, genomics, and machine learning.