Annotated Cell-Line Resources Speed Discovery

Two groups screen hundreds of cancer cell lines to find biomarkers of drug sensitivity

As automated “omics” technologies transform biomedical research, cancer biologists are rapidly taking advantage of the massive screens enabled by these techniques. Two collaborative groups have each screened several hundred human cancer cell lines, creating databases of the cell lines' genetic characteristics and responses to multiple anticancer drugs. The goal of these technologies is to identify characteristics that predict the cells' response to different agents, providing biomarkers both to speed cancer drug discovery and, eventually, after additional preclinical and clinical validation, help oncologists select the best treatment for patients.

The 2 studies, which both published their initial results on March 28 in Nature, are expansions of an effort pioneered by the U.S. National Cancer Institute with its NCI60 cell line panel and associated drug screens.

The Cancer Cell Line Encyclopedia (CCLE) compiles gene expression, chromosomal copy number, and massively parallel sequencing data on 947 cell lines, as well as pharmacologic profiles for 24 anticancer drugs screened across 479 of the cell lines. The collaboration included researchers from the Broad Institute of MIT and Harvard (Cambridge, MA), Dana-Farber Cancer Institute (Boston, MA), Harvard Medical School (Boston, MA), the Novartis Institutes for Biomedical Research (Cambridge, MA), and Howard Hughes Medical Institute (Chevy Chase, MD). This work was supported primarily by a grant from Novartis.

The team analyzed these data using 2 predictive algorithms constructed with machine learning techniques and identified many known genetic, lineage, and gene-expression–based predictors of drug sensitivity. The researchers also found 3 new correlations among cell lineage, gene expression, and drug efficacy, suggesting that such large, annotated cell-line collections could help preclinical researchers determine which agents are most likely to help with specific cancers.

“It not only speeds things up, but it allows you to consider experiments you couldn't have contemplated before,” says Todd Golub, MD, chief scientific officer of the Broad Institute and a corresponding author of the paper. “It will have a profound effect on the types of preclinical studies people can do in oncology. It will resurrect the value of cancer cell lines.”

As an example of the sort of experiment the CCLE data could enable, Golub suggests that researchers could screen a panel of cell lines with a library of small molecules and correlate the results with the genetic features of the lines listed in the encyclopedia. “Previously one could observe, oh well, some small molecules kill certain cell lines and not others, but you'd have to throw your hands up and say, gee, I wonder what that's all about,” says Golub. “Now we can go to the Cell Line Encyclopedia and look up what the molecular basis of that might be.”

The CCLE is freely available for researchers, academic or corporate, to consult for their research. Golub says the team plans to increase its annotation of the cell lines in the collection, including proteomic and metabolic data, along with functional perturbations such as RNA interference screens.

A second collaboration, the Cancer Genome Project's Cancer Cell Line Project, screened 639 tumor cell lines that included both common and rare childhood and adult cancers. The researchers sequenced the full coding exons of 64 commonly mutated cancer genes, performed genome-wide analyses of copy number gain and loss, and profiled the expression of 14,500 genes in each line. They then screened 130 drugs across various cell lines for a total of 48,178 drug–cell-line combinations.

In its Nature article, this team also reported both known and unexpected relationships between genetic factors and drug sensitivity. The collaboration included researchers from the Wellcome Trust Sanger Institute (Cambridge, UK), Massachusetts General Hospital, Dana-Farber Cancer Institute, Harvard Medical School, European Bioinformatics Institute (Cambridge, UK), Institut Curie (Paris, France), Centre Hospitalier Universitaire Vaudois (Lausanne, Switzerland), and the Howard Hughes Medical Institute. Its data are also available as a resource for all cancer researchers.