With CellMiner, Researchers Prospect for Cancer Discoveries
Researchers have updated a web-based application that makes a wealth of genomic and pharmacologic data obtained from a panel of human cancer cell lines available to anyone with access to a computer and the Internet.
Known as CellMiner, the publicly available application lets researchers rapidly retrieve data on the expression of more than 22,000 genes and 360 microRNAs, and the growth-inhibiting activity of more than 26,000 compounds in NCI's Human Tumor Cell Line Screen, also known as the NCI-60 panel of cell lines. The compounds include 102 FDA-approved drugs, as well as others being studied in clinical trials.
Tapping into a Wealth of Information
The NCI-60 consists of 60 human cancer cell lines (laboratory-grown cells) derived from nine different tissues. These cell lines were acquired by NCI's Developmental Therapeutics Program (DTP) to screen compounds for anticancer activity. In addition to testing about 100,000 compounds on these cells, NCI researchers have amassed an extensive database of genomic information for these widely used cell lines.
A Case Study of CellMiner-Aided Discovery
A study led by Dr. Gabriele Zoppoli of NCI's Laboratory of Molecular Pharmacology and published September 11 in the Proceedings of the National Academy of Sciences, exemplifies the discoveries that can be made with CellMiner.
The researchers used the tool to search for genes whose expression correlates with the cell-killing activity of anticancer drugs called topoisomerase inhibitors in the NCI-60.
"At the very top of the list there was one gene, which we knew nothing about, called Schlafen-11 (SLFN11)," Dr. Pommier said.
Pursuing this lead, the researchers found that SLFN11 expression is causally associated with the activity of a broad and widely used group of chemotherapy drugs known as DNA-damaging agents (DDAs). In a series of experiments they showed that the gene sensitizes cancer cells to DDAs, has a wide expression range in colon and ovarian cancer samples from The Cancer Genome Atlas, and may act as a biomarker for predicting the response to DDAs in patients.
"No one had heard of this gene before or knew anything about it" until it turned up in the CellMiner results, noted Reinhold, who was a co-author on the study. "It shows how you can discover things [with CellMiner] that you could never have anticipated."
Working with such large data sets often means dealing with unwieldy databases that make it difficult to analyze and integrate data. But with CellMiner, researchers can easily compare patterns of gene expression, microRNA expression, and drug activity in the NCI-60. By means of a novel pattern-matching tool, users can also explore relationships between these parameters and any pattern of interest they choose to input (for example, cell lines that lack mutations in the commonly mutated tumor suppressor gene TP53), enabling them to define their own questions.
Such pattern comparisons can, for example, reveal new connections between drug activity and gene expression, pick out compounds or drugs that work through similar—or complementary—mechanisms, or identify genes that may predict the response of cancer cells to specific drugs, noted Dr. Yves Pommier, chief of the Laboratory of Molecular Pharmacology (LMP) in NCI's Center for Cancer Research (CCR).
No Bioinformatics Expertise Needed
"Our goal is to have this database used by people who do not have bioinformatics expertise, including M.D.s and anyone else who wants to explore the database without having a bioinformatics team next to them," said Dr. Pommier, who co-authored a recent report in Cancer Research that detailed CellMiner's features and provided case examples of its use. The lead author and lead developer of CellMiner, William Reinhold, is a molecular biologist in LMP.
Reinhold, Dr. Pommier, and their colleagues developed CellMiner so that researchers could avoid the time-consuming data processing previously required to work with information in the NCI-60 database. The suite of web-based tools provides "a quick and easy way for people to start doing systems biology and pharmacology—which is to say, comparing big data sets of disparate types to ask scientific questions," explained Reinhold.
"Without specialized expertise, there's been a huge wall between people who want to ask these questions and the people who have access to the information," he continued. "We're trying to take that wall down."
Users of CellMiner simply input their query online and, within minutes, receive an e-mail containing the results in the form of tables and bar graphs in a single Excel spreadsheet. The application calculates the correlation between all parameters and identifies statistically significant correlations.
And, because data are provided in a spreadsheet, Dr. Pommier noted, users can archive the results and continue working on them, even without Internet access. Researchers "can use Excel tools to search and organize the data, which makes it very versatile," he added.
"CellMiner is a powerful tool that allows you to generate hypotheses about the ways in which different genes or patterns of gene expression can affect cell behavior [in cancer]," said Dr. Michael Gottesman, chief of CCR's Laboratory of Cell Biology. His lab, which was not involved in developing CellMiner, is using the tool to study gene expression patterns that correlate with resistance to specific anticancer drugs.
Investigators in Dr. Susan Bates's lab, in CCR's Molecular Oncology Branch, were among the first to use CellMiner's predecessor—a computer program called COMPARE—to probe the NCI drug-screen database. Using COMPARE, they identified a drug with anticancer activity called romidepsin, which turned out to be effective against T-cell lymphoma.
Because only one-third of patients with T-cell lymphoma responded to romidepsin in clinical trials, Dr. Bates's team is using CellMiner to help find a biomarker that predicts which patients will respond favorably to the drug.
"The CellMiner program builds on COMPARE by providing a much easier-to-use interface," Dr. Bates said. "People in my laboratory…have found it very easy to adapt to CellMiner."
Working with CellMiner "is one-stop shopping, whereas, before, working with the drug screen data required multiple queries," confirmed Dr. Robert Robey, a chemist in Dr. Bates's lab who works with the tool. "I read the paper and within a few minutes I was pulling up drug profiles. The interface makes it easy to put in your own data and pull meaningful things out of it."
"Users of CellMiner need to realize that the COMPARE program on the DTP website still represents the gold standard for determining correlations using drug profiles," Dr. Bates noted. "CellMiner provides an easy-to-use format. But there are limitations in terms of the results it is able to generate. There is a larger body of data available on the DTP website, both for compounds and for molecular characterization data."
CellMiner's developers plan to continue updating and enhancing the software. The next version will provide access to the whole-genome sequences for all protein-coding regions, or exons, of the genome across the NCI-60, Dr. Pommier indicated. The team is also making the open-source software available for others to use or modify as they wish to incorporate molecular profile data on other cell lines or human tumor samples.
Without specialized expertise, there's been a huge wall between people who want to ask these questions and the people who have access to the information. We're trying to take that wall down.
—Dr. Yves Pommier
Cell lines from the NCI-60 panel, which are studied by cancer researchers worldwide, "have formed the basis for a lot of what scientists know about the physiology of cancer," Dr. Gottesman noted. But, as reported last November, research led by Dr. Jean-Pierre Gillet in Dr. Gottesman's lab has suggested that these and other cancer cell lines may have important limitations when used to identify genes associated with resistance to chemotherapy drugs in particular tumor types. The study showed that, in various cancer types, the expression of a specific set of genes associated with drug resistance was very different in cell lines from what it was in tumor samples representing the same cancer types.
The fact that some gene expression patterns in laboratory-grown cells such as the NCI-60 differ from those in the original tumor tissue is not surprising, researchers agree. However, Dr. Pommier said, "there is a vast array of genes in the NCI-60 that retain their expression pattern between the cell line and the tumor—that's what is coming out of the CellMiner analyses, because now it's easy to look at those genes."
Although researchers continue to debate the extent to which cancer cell lines represent the tumors from which they originated, "there's a lot we can still learn from the NCI-60," Dr. Gottesman said, "and having a tool that allows you to gain access to the huge amount of data that's been accumulated is a very useful thing."
The CellMiner project was supported by NCI's Intramural Research Program.