AdaptML: Modeling the evolution and ecology of gene families

To date, modeling the behavior of cellular networks under laboratory conditions has received more attention than modeling how ecological factors affect diversity in natural environments. As we move toward the ultimate goal of integrating laboratory model organism studies with field data, a key challenge will be identifying the geochemical/ecologial factors that underlie community diversity, and the phylogenetic boundaries of natural ecological populations. Thus, computational frameworks for automatically learning models of sequence evolution in the context of metadata (e.g., site geochemistry/ecology) will need to be developed.

We have developed one such framework: AdaptML, a maximum-likelihood-based tool for studying both the sequence evolution and ecological history of a set of gene sequences. To perform this latter task, AdaptML employs a hidden-Markov-model-like strategy of assigning gene sequences to unseen states we term "habitats." These habitats are inferred automatically and designed to recapitulate sequence partitioning observed in the wild. AdaptML was initially developed and tested in collaboration with Martin Polz’’s lab, using data from 1027 strains of marine Vibrio hsp60 gene sequences harvested off the coast of Maine. We showed that AdaptML can be used to analyze this dataset and to help build models of Vibrio resource partitioning.1

References

1. Hunt, D. E.*, David L.A*, et al. Resource Partitioning and Sympatric Differentiation Among Closely Related Bacterioplankton. In press, Science (New York, NY) (2008).

AdaptML: Modeling the evolution and ecology of gene families

5/10/08

Identifying the geochemical/ecological factors that underlie community diversity, and the phylogenetic boundaries of natural ecological populations.

Run/download AdaptML here

Figure 1. Circular phylogenetic tree of marine bacterial hsp60 sequences isolated in the Gulf of Maine. Encoded in the outermost ring is information regarding each sequences’ sampling period: [green=spring; red=fall]. The innermost ring depicts bacterial retrieval location: [yellow=living zooplankton; red=dead zooplankton; green=grassy material; light-blue=globular non-zooplankton material; dark blue=other]. Interior circles designate microhabitat assignments learned by AdaptML. Sequences associated with green microhabitats are grouped by virtue of being associated with the fall; both red and purple microhabitats reflect association with the spring. But, red microhabitat-associated isolates are more likely to be retrieved from zooplankton than purple-associated isolates. (Raw data provided courtesy of Sarah Preheim and Martin Polz.)

Next >

< Previous