Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Data Mining Knowl Discov
Impact Factor: 2.541

Mining high‐throughput screens for cancer drug targets—lessons from yeast chemical‐genomic profiling and synthetic lethality

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract The recent decrease in the rate that new cancer therapies are being translated into clinical use is mainly due to the lack of therapeutic efficacy and clinical safety or toxicology of the candidate drug compounds. An important prerequisite for the development of safe and effective chemical compounds is the identification of their cellular targets. High‐throughput screening is increasingly being used to test new drug compounds and to infer their cellular targets, but these quantitative screens result in high‐dimensional datasets with many inherent sources of noise. We review here the state‐of‐the‐art statistical scoring approaches used in the prediction of drug–target interactions, and illustrate their operation using publicly available data from yeast chemical‐genomic profiling studies. The real data examples underscore the need for developing more advanced data mining approaches for extracting the full information from the high‐throughput screens. A particular medical application stems from the concept of synthetic lethality in cancer and how it could open up new opportunities for personalized cancer therapies. © 2012 Wiley Periodicals, Inc. This article is categorized under: Algorithmic Development > Biological Data Mining Application Areas > Health Care

The concept of synthetic lethal genetic interactions and its use in high‐throughput screening of drug targets. (a) Systematic mapping of genetic interactions between genes (lines) provides functional insights into the organization of cellular networks by revealing compensatory pathways (arrows) involved in essential cell functions (e.g., proliferation). (b) Synthetic lethality is an extreme case of genetic interactions, in which perturbation of either gene alone through genetic mutation (X), RNA interference (RNAi) silencing or drug treatment is nonlethal, but simultaneous mutation of both genes results in lethality. (c) Selective drugs or targets are those chemical compounds or RNAi reagents, respectively (black boxes), which kill only those cells that harbor specific mutations while sparing normal cells. Accurate prioritization of the most promising hits in high‐throughput assays relies on effective mining of the resulting high‐dimensional data matrices through statistical approaches.

[ Normal View | Magnified View ]

Comparative heat map of the three scoring approaches. Only the two example chemicals and 420 top‐ranking genes are shown. The numbers in the cells indicate the rank obtained using the particular approach, with ranks varying from 1 to 992 within both chemicals. The darker the cell color, the larger the value of the scoring function. The interacting genes obtained from STITCH are bolded; genes labeled with one and two asterisks indicate the chemical–gene interactions for rapamycin and benomyl, respectively. The gray cells indicate missing values.

[ Normal View | Magnified View ]

The two example chemicals used to illustrate the different drug target identification approaches: (a) rapamycin. (b) benomyl. For both chemicals, illustrations include its structural formula, chemical–gene interaction network and the ranks of the known or predicted interacting genes using different scoring approaches. The interaction networks were retrieved from the STITCH database using cutoff 0.700 (high confidence). The red tablets stand for chemicals and spheres for genes. Green lines indicate chemical–gene interactions and blue lines protein–protein interactions between the corresponding protein targets. The thicker the line, the stronger the confidence of the interaction. In the rank lists, only the top‐ranking genes with STITCH link to the particular chemical are shown. The color of the gene indicates the scoring approach used in the ranking. It should be noted that STITCH may retrieve also genes which are not direct targets of the chemical under analysis; for instance, AVO2 and BIT61 belong to the TORC2 protein complex, but are not sensitive to rapamycin. However, their interaction partners TOR1 and TCO89 are sensitive and known targets of rapamycin.39 All associated genes retrieved from STITCH are shown here for completeness.

[ Normal View | Magnified View ]

Distribution of the rankings of the STITCH chemical–gene links for the three scoring approaches. Out of the 327 chemicals involved in the chemical–genomic datasets, 89 chemicals had at least one STITCH link. The bars summarize the number of those chemicals for which the rank of the best ranking STITCH link belongs to the prespecified ranges when using the different scoring approaches.

[ Normal View | Magnified View ]

Heat map illustrations of selected, small proportions of the three scoring matrices. (a) Fitness defects in the chemical‐genomic screening (FD score). (b) Genetic interaction scores in the SGA dataset (ε‐score). (c) Correlation coefficients between the chemical–genomic and genetic interaction profiles (ρ‐score). The gray cells indicate missing values. The white boxes with labels a–e correspond to the cases discussed in the main text.

[ Normal View | Magnified View ]

Browse by Topic

Application Areas > Health Care
Algorithmic Development > Biological Data Mining

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts