Global analysis of gene function in yeast by quantitative phenotypic profiling

Abstract

We present a method for the global analysis of the function of genes in budding yeast based on hierarchical clustering of the quantitative sensitivity profiles of the 4756 strains with individual homozygous deletion of nonessential genes to a broad range of cytotoxic or cytostatic agents. This method is superior to other global methods of identifying the function of genes involved in the various DNA repair and damage checkpoint pathways as well as other interrogated functions. Analysis of the phenotypic profiles of the 51 diverse treatments places a total of 860 genes of unknown function in clusters with genes of known function. We demonstrate that this can not only identify the function of unknown genes but can also suggest the mechanism of action of the agents used. This method will be useful when used alone and in conjunction with other global approaches to identify gene function in yeast.

Figure 1

Two-way unsupervised uncentered unnormalized hierarchical clustering using a Pearson's correlation of the phenotypic profiles of 4281 nonessential genes to 51 different treatments. The expanded region shows the DNA-damage cluster, which contains the components of the DNA-damage checkpoint function, nucleotide excision repair, and homologous recombination.

Figure 2

Precision–recall evaluation of phenotype data on GO biological processes. The predictive power of phenotype profile correlations was evaluated against a gold standard based on six biological processes as defined by the GO: DNA repair, amino-acid biosynthesis, cell cycle checkpoint, response to osmotic stress, aerobic respiration, and galactose metabolism (A). The fraction of known functionally related gene pairs to total predictions (precision) at a range of thresholds is plotted versus the percentage of the number of known gene relationships recovered (recall) (). The characteristics of other high-throughput experimental data, affinity precipitation (▪), yeast two hybrid (), synthetic lethality (), transcription factor binding site data (), microarray correlation (), and functional data derived from Hughes et al (2000) () are shown for comparison. Two supervised feature selection methods were used to select the relevant features from the diverse collection of microarray data, one selecting single data set features independently and the other including or excluding entire data sets. The phenotype data is both more sensitive and precise than other high-throughput data on this set of processes. The phenotype profiles were also evaluated against a more general set of GO terms for comparison against existing data including (B) and excluding (C) the ribosome biogenesis GO term (GO:0007046), which tends to dominate gene pairs implicated by coexpression. The phenotype profiles implicate gene relationships over a broad range of biological processes.