Editors' ChoiceCancer

Meta-screen for cancer dependencies

+ See all authors and affiliations

Science Translational Medicine  09 Aug 2017:
Vol. 9, Issue 402, eaao2262
DOI: 10.1126/scitranslmed.aao2262


Systematic identification of gene dependencies in cancer cells teaches us some lessons.

What is the most useful information we can obtain about a cancer in order to target it therapeutically? Researchers have developed a multitude of analytical tools to endlessly identify aberrations in genes, proteins, and even entire pathways in cancer cells. However, such efforts leave us with an incomplete picture because many features may simply be associated with certain cancers but not vital for their survival. To address this, Tsherniak, Vazquez, and co-workers used shRNA-based screens to systematically identify genetic dependencies in over 500 cancer cell lines. shRNAs are notorious for having off-target effects but, refreshingly, the authors directly addressed this by developing a computational method to correct for them. After correction, the authors found 769 strong dependencies within the cell lines, 152 of which are categorized as druggable. In addition, 53 dependencies were identified in at least 5% of the cell lines. They also looked for biomarkers associated with the dependencies and, surprisingly, found that the vast majority (82%) of dependencies could be predicted by gene expression, not copy number or mutational status. This suggests that personalized medicine efforts may be more effective using transcriptional profiling rather than tumor sequencing.

The scale of screening and analysis undertaken in these studies is impressive and doubles the magnitude of previous efforts. However, the authors report that similar studies across over 5000 cancer cell lines are needed to exhaustively identify strong dependencies and their associated predictive features. Although such an effort would undoubtedly yield new targets and biomarkers, they would be associated with fewer and fewer cancer cell lines, and thus the value of each identified dependency would ostensibly diminish. Instead, patients may benefit more from the application of the technology and analyses used here in alternative screens yielding insights with broader implications and greater clinical potential. For instance, by directly screening primary patient tumors, which could be accomplished on a more modest scale, researchers may be able to identify actionable dependencies that are present in patients. The same can be said of screening in vivo mouse models, which may offer insights into tumor-host interactions and yield entirely new classes of targets for therapeutic development.

Understanding which genes are vital within or across cancers of different lineages will help researchers prioritize their drug development efforts. Because of this, the cross-cancer meta-analysis is a particularly notable feature of the dataset presented in this study (available at depmap.org).

Highlighted Article

View Abstract

Navigate This Article