All the versions of this article:
Can the results of high-throughput phenotyping experiments like those performed in the context of the Metabolic Thesaurus project be exploited systematically in yeast as well, and to a lesser extent in E. coli, in order to obtain functional information which is more elaborate than that obtained in specific genetic experiments?
In fact, these experiments make it possible to associate a phenotypic profile with each gene, i.e. the list of growth phenotypes of the corresponding mutant on all the media tested. To a certain extent, the similarity between the phenotypic profiles of two genes reflect the proximity of the roles of these genes in the mechanisms which allow the bacterium to survive on these media. The clustering of genes with similar profiles can therefore provide strategies for the annotation of genes of unknown function, for example, by relating them to other genes of known function.
The first step consists of defining one or more methods of classification (clustering, biclustering) adapted to this type of data, calibrating them, and evaluating their pertinence relative to the ensembles of genes of known function.
For all methods of classification used, the functional interpretation of the groups of genes obtained is complicated by the fact that only a small fraction of the genes already have an annotation. A possible strategy is to complete the information contained in the phenotypic profiles using data on functional relationships between genes either measured or inferred from a different type of experimental data (* Von Mering et al (2005)). The groups of genes may then be analyzed in the context of a network constituted of functional links which permits the elaboration of hypotheses for gene function from an extended ensemble of “neighboring” genes. Finally, do the results of high throughput phenotyping enable better understanding of the link between cellular phenotypes and molecular mechanisms? In fact, these results correspond to a systematic double perturbation of these mechanisms: deletion of a gene and changes in environmental conditions. It is possible to interpret them in terms of a network of metabolic reactions, sometimes complemented by transcriptional regulation links? With this objective we are studying the link between the metabolic network and the distances between phenotypic profiles.
[*] Von Mering C, Jensen, Lars J, Snel, Berend, Hooper, Sean D, Krupp M, Foglierini M, Jouffre N, Martijn A, Bork P.
“STRING : known and predicted protein-protein associations, Integrated and transferred across organisms.”
Nucl. Acid Res.2005. (33): D433-437.