We have developed a genomic approach to target identification in yeast that may be broadly applicable to human therapeutics. This approach is based on our observation that drug targets can be identified by their ability to confer sensitivity to a cell when present at a reduced copy number; in this a case change in copy number from two copies in a diploid yeast strain to one copy in a heterozygous deletion strain. We have assessed the feasibility of this approach by confirming the molecular targets of several currently available drugs (Giaever et al. 1999). Briefly, in our method, a complete genome set of molecularly bar-coded heterozygous yeast strains are pooled, grown competitively in drug and analyzed for relative growth rates using high-density oligonucleotide arrays. Growth rate is a metric of relative sensitivity of each strain. The strain most sensitive to drug in many cases identifies the gene encoding the drug target. In this way, each strain represents a potential drug target and thus the complete set of potential targets is screened in a single assay. This method has the advantage of generating targets in an unbiased manner, as it is the organism itself that reports its most essential targets in the presence of any given compound. For this reason, we expect that we will uncover novel molecular targets and compounds in a manner that would up-end the current drug discovery paradigm. Instead of the subjective approach of pre-selecting targets, we aim to simultaneously identify and validate novel, essential molecular targets and the compounds that inhibit them in the """"""""genomically"""""""" accessible yeast S. cerevisiae. Given that approximately 50% of all known yeast genes have human homologs (Foury 1997), it is our contention that the identification of all gene products essential for yeast growth will be relevant to human cancer. In addition, because each strain can be ranked in order of sensitivity, sensitive strains other than the strain that identifies the drug target will assist in the deconvolution of the involved pathways. These pathways include the drug target pathway as well as drug response pathways. Through the generation of a database containing thousands of profiles of known drugs and unknown compounds, it may be possible to gain insight into mechanisms of drug action and compound structural similarities through the clustering of compound profiles. Furthermore, because our technology defines a novel genome= wide measurement, we believe it will significantly contribute to the elucidation of gene function ultimately arising from the combined analysis of genomic data sets collected using diverse genomic technologies.
