Genome Technology Center

Haplo-insufficiency project: automated quality scoring


Guri Giaever
Jochen Kumm
Patrick Flaherty
Adam Arkin
Dan Jaramillo
Michael Proctor

Related Info

Yeast screen robot for HIP

Molecular tags arrayed on the Affymetrix TAG3 Chip can be used to score the relative abundance of pools of all single gene knockout strains in yeast. A large data set of experiments scoring genome-wide changes in the presence of drugs affords insight into mechanism of action and synergistic interactions of drugs.

Many experiments are performed with different drugs and over extended periods of time. Strain-by-strain comparisons of relative fitness - measured by intensity of strain specific tags on a TAG3 chip – can be used to identify both low quality data points and drug that affect many genes. Using rank-order based statistics and a distribution created by permutations, detection and classification of outliers has been automated. This allows for consistent quality control and improved confidence in analysis results.

Gene knock-outs have been a crucial tool for geneticists and have yielded much insight into gene function. Generally a knock-out study focuses on a single gene or small group of genes identified in a phenotypic screen, and this gene-by-gene approach is straightforward in organisms such as yeast and flies. It is difficult in most organisms, and gene knock-outs only allow analysis of non-essential genes.

Fast acting, reversible chemical probes are another way to perturb the function of essential genes. The main objective of the work proposed is to develop bioinformatics tools to process and analyze data from a large set of experiments measuring the effects chemical probes in the yeast proteome. Using a genome-wide assay allows us to profile the relative sensitivity of all proteins targeted by a chemical probe in a cell. This chemogenomic assay is based on the finding that removing (knocking-out) one of the two copies of a gene in diploid yeast results in a heterozygous cell sensitive to a chemical probe acting on its gene product (Nat Genet 21:278, 1999). A collection of 5,918 heterozygous deletion strains is pooled and grown competitively in the presence of a chemical probe. Molecular bar-codes uniquely identify each strain and the relative fitness of each strain in the pool is measured by oligonucleotide array hybridization.

The assay has screened 1,200 known compounds showing that a strain heterozygous for a known target is highly sensitized to the compound. Breakthroughs in miniaturization and automation now enable high-throughput screening of large numbers of compounds. We propose building a searchable repository of ~10,000 compound-genome interactions. To do this, algorithms and tools must be scaled by an order of magnitude and data analysis must be automated. As part of this we propose to characterize compound-genome interactions computationally. Since this assay detects direct connections between compounds and the yeast proteome, we can build a network of functional interactions in yeast from this data set This provides significant novel information about the interconnectivity of the proteome extending data sets derived from Y2H assays. We also propose to create a chemogenomic map of the yeast proteome comparing existing ontologies for chemical probes to measured functional effects enabling inference of pathways by measuring the covariance of functional effect and compound classification. We will also propose de novo classifications of compounds based on functional effects alone.

Such genome-wide exploration of biological and chemical space potentially identify novel drug targets and allow characterization of chemical structures that comprise "drug-like" compounds. The proposed repository of the functional genomics of chemical probes will be significant resource for understanding small molecule effects in the cell.

Footer Links: