This Metabolic Gene Card gives an overview of the metabolic profile upon deletion of the gene KES1/YPL145C. It further lists other genes that have a similar metabolic impact, and hence, likely have a similar function. Further, we provide an enrichment analysis to dissect the obtained functional clusters, and compare the obtained information with physical and genetic interaction networks.
The gene card provides a summary about the response of the prototrophic Saccharomyces cerevisiae deletion strain KES1/YPL145C under exponential growth in minimum synthetic medium - a condition were biosynthetic metabolism is expected to be highly active. The concentration of the free amino acid profile is a direct phenotypic readout for the gene deletion and allows us to estimate the genetic contribution of each of 4678 tested genes towards amino acid homeostasis.
Genes with similar function are expected to cause a similar metabolic signature. We assign metabolically similar strains to 280 distinct groups using hierarchical clustering and test them for overrepresentation of gene ontology (GO) functional terms, phenotypes and co-citation in literature. Thereby, we can suggest a gene function for not well characterised genes.
Amino acid levels are given in relation to the average of 4678 deletion strains. Robust estimates for mean, standard deviation, and covariance for each amino acid, and the Mahalanobis distance as metric for the overall change on the amino acid profile, were calculated using the Minimum Covariance Determinant (MCD) (R package: robustbase). Accordingly, we determined significance of change in the univariate case with a Z-test and in the multivariate statistic with a χ2-test statistic. All p values were adjusted for multiple testing with the FDR method from Benjamini & Hochberg (p.adjust function in R).
Identification of 1519 altered amino acid profiles, 1459 by multivariate (χ2-test) and 1069 by univariate statistics (Z-test). Deletion strain KES1/YPL145C is highlighted.
The amino acids are variance scaled (z-scores) and ranked from low to high for easier interpretation. Significance of change is indicated in color and values given in table below.
to the top
Similar metabolic profiles are indicative for a common genetic role in the regulation of amino acid metabolism. Furthermore, they are informative about functional relationship and therefore valuable for hypothesis generation and for functional characterisation of not well characterised open reading frames.
Groups of strains with similar amino acid profiles were identified by consensus clustering using 500 repetitions of hierarchical clustering based on pairwise Mahalanobis distances upon perturbation of the data. The obtained consensus matrix, which records the number of times two strains were assigned to the same cluster, was substracted from 1 and used as distance metric to identify 280 metabolically distinct and robust clusters (R package dynamicTreeCut).
Illustrated are the concentration values in a heatmap for deletion strains with a similar amino acid profile to KES1/YPL145C. The dendrogram was calculated using Wards Criterion as agglomeration method. Concentration values are variance scaled (z-scores) and magnitude of change indicated in color.
To summarise the available data for the set of metabolically similar strains we construct a network based on genetic and physical interactions obtained from the BIOGRID database using the freely available esyn toolkit (2014 Bean et al). The online tool organises the information from BIOGRID and adds several more functionalities to analyse the network. It allows to retrieve information for each interaction, add common interactors, manually add or delete genes, perform network analysis tasks, or integrate data from Yeastmine.
Enrichment analysis is a further tool to put new findings into the context of established knowledge. In order to find predictive enrichments, we remove the strain of interest from the cluster and the background set before the enrichment analysis. We use the gene ontology definition (GO Consortium), KEGG annotation and the phenotype and literature database from SGD. Enrichment analysis for GO annotations was performed in R with the GOstats package. In this package p values were not adjusted for multiple testing.
Enrichment analysis for phenotypes and literature annotations was done using hypergeometric testing. Obtained p values were corrected with the FDR method from Benjamini & Hochberg (p.adjust function in R).
Enrichment analysis based on metabolic signature for:
Mülleder, M., Calvani, E., Alam, M.T., Wang, R.K., Eckerstorfer, F., Zelezniak, A., and Ralser, M. (2016). Functional Metabolomics Describes the Yeast Biosynthetic Regulome. Cell. dx.doi.org/10.1016/j.cell.2016.09.007
Information on gene ontology terms, KEGG terms, phenotype and co-citations was retrieved on 2016-08-10.