利用微生物培养基的景观来预测新的有机体媒体配对

The most frequently altered compounds in this way are the biologically common ions/salts, followed by trace metals and vitamins. This lends further evidence that these trace components play key roles in differentiating growth between close species, and thus should be considered in future media design.

Beyond these analyses, we examine broad trends in compound usage across phyla at different taxonomic levels. Heat maps of enrichment of different taxonomic groups for media components can be found in Supplementary Figs 4–8 and Supplementary Note 3.

Media usage follows phylogenetic and ecological trends

An implicit assumption that investigators make when trying to cultivate new microorganisms is that the best medium to start with is one from a phylogenetic or ecological neighbour. Despite its apparent logic, this assumption has not, to our knowledge, been rigorously tested and validated. To do this, we mapped organisms in DSMZ to operational taxonomic units in Greengenes ecological data as clustered into environments (see Methods for details; clustering in ref. 20), and also to taxonomic classifications from the Interactive Tree of Life project (Itol21). We find that, indeed, the likelihood that two organisms share at least one lab medium strongly correlates with both their ecological and phylogenetic similarity (see Fig. 3; ρ=0.76, P=2.3e?13, and ρ=0.92, P=1.3e?3, respectively, for ecological and phylogenetic similarities, as determined by cohabitation Jaccard index (ecological) or inverse subtree count in the iTOL taxonomic tree (phylogenetic); see Methods for details). This indicates that phylogenetic and ecological closeness are good heuristics for determining the likelihood that two organisms have successfully been grown in the same lab medium. Indeed, we show later that this is not just descriptive of what has been done in the past, but that it holds a signal that can be used predictively for deriving new successful organism–media pairings. Importantly, the fractions of organism pairs sharing lab media listed in Fig. 3 are likely underestimates, as the organism-by-media matrix in KOMODO is highly underpopulated (see previous section). This observation is indeed upheld when we perform new growth experiments, as most of our predictions (which were not listed previously in KOMODO) yield growth.

文章引自：nature.com；版权声明：版权归原作者所有，如有版权问题，请与我们联系。