The hottest recent trend in cancer research is to think big—very big.

The goal of the $100 million-a-year Cancer Genome Atlas (TCGA) project is to perform a dazzling array of genetic tests on thousands of tumors from thousands of patients with a range of cancers. These tests start with the sequencing of the gene-encoding regions of tumors (“whole exome sequencing”). They often move to the sequencing of entire tumor genomes (“whole genome sequencing”)—among others.

The result has been a mind-boggling number of new cancer data streams and potential drug targets. They are letting researchers zero in with historic accuracy on the unique, if complex, genetic signatures of each patient’s tumors. Countless new cancer subtypes are being unearthed.

The new data are also leading to the discovery that certain cancer subtypes are more like subtypes of other cancers than their own.

The popularity of Big Data projects was highly evident at the April 6-10 American Association of Cancer Research (AACR) meeting, whose theme was “personalizing cancer care through discovery science.” Session after session featuring TCGA was Standing Room Only (SRO). Washington Convention Center attendants struggled to keep up.

One session was so jammed, TCGA program office director Kenna Shaw stopped her talk to ask the crowd to tell AACR organizers about it, so TCGA would get bigger rooms next time. TCGA sessions were overflowing last year too, she said.

Many sessions featured discussions of a stunning array of new analytical TCGA tools. Others featured a stunning array of new insights being formed by those tools.

Harvard genetics professor Raju Kucherlapati, principal investigator of a TCGA Genome Characterization Center, offered a sample of the “really, really interesting” findings his group has made studying chromosomal rearrangements in many cancers, including colon and endometrial.

In the talk, “Structural Aberrations in a Thousand Tumors,” he noted it has recently become known high-frequency mutation rates in colon cancer, and endometrial cancer subsets, counter-intuitively correlate with “exceedingly good prognosis.” TCGS colleague Chris Sander, Memorial Sloan Kettering Computational Biology Program Chair, surmised this may be because some tumors replicate so fast they “wear themselves out.”

Regardless, a new way to predict outcome, target drugs, and gauge regression of these tumors is to look for mutation frequency, Kucherlapati said. But TCGA work reveals the common culprit behind high-frequency mutations in colon and endometrial cancers--microsatellite instability-- is not always to blame. Many tumors with high-frequency mutation rates studied by TCGA did not express microsatellite instability.

What all did seem to share were mutations in the proofreading domain of polymerase epsilon. So colon cancer experts should seek a more thorough way to gauge mutation rate, and generate more targeted treatments, he said.

Another area that may be affected by TCGA: immunotherapy. Kucherlapati and others find cancer-related chromosomal rearrangements tend to cluster in limited areas of tumor genomes, including HLA (histocompatibility antigen) DNA regions.

“This suggests the ability of cells to present antigen to the immune system is abrogated in cancer, and might be a mechanism by which tumors escape immune surveillance,” he said. Translation: certain mutations in cancer cells render them invisible to immune cell attack. An earlier TCGA report found similar HLA rearrangements in squamous cell lung cancers.

Kucherlapati’s group further discovered many virus-induced cancers are not just driven by transforming viral genes, as was thought. (Many cancers can be initiated by viruses, including cervical, head and neck, and bladder cancers.) Viruses can also insert themselves into, and disrupt, genome regions regulating tumor suppression. So while vaccines successfully tackle certain virus-induced cancers, coupling vaccines with approaches targeting errant tumor suppressors may help eradicate some cancers in parts of the world.

One bottom line, said Kucherlapati: “If you want to really develop personalized therapies, you need to be looking at mutations in a large number of genes, not just a few canonical genes. There are many different pathways, and many different ways a tumor can go awry.”

In fact, Sander noted the new flood of individualized data requires “a new kind of clinical trial.” Since TCGA is detecting far more subsets of cancers, each experimental drug will have to be tried on smaller patient populations in any given area. Some cancer centers are already forming networks to accommodate this.

The free exchange of “pre-competitive data” to build a massive new “knowledge commons” is key, said Sander, echoing the plenary address of National Cancer Institute (NCI) chief Harold Varmus.

Furthermore, MD Anderson Cancer Center Systems Biology chief Gordon Mills was one of many providing evidence that, in the face of Big Data, genome analysis groups will need to rely on a great variety of cancer experts.

His group is generating for TCGA a functional proteomics atlas of all its tumor types. They developed a reverse-phase protein array system to look at tumor samples. (“Think of it as a high-throughput, multiplex, inexpensive Elisa test,” he said.) After running 4,000 tumor samples, many clusters “jumped out.” Integrating his data with that of collaborators like Sloan Kettering oncologist Douglas Levine, he has found many bladder cancers are marked by the same human epidermal growth factor receptor 2 (Her2) mutation driving many treatable breast cancers.

So some bladder cancer patients may receive effective breast cancer drugs, like Herceptin.

Integrated TCGA work also finds the androgen receptor to be “the best single marker in the world for breast cancer,” said Mills. “I am not sure many of you would have predicted this. It is an incredible marker of luminal breast cancer.”

Big Data may ally experts not only working on cancer with different tools, but different philosophies. Geneticists believe cancer is a genetic disease. But immunologists think of it as more of an “immunologic disorder,” as Johns Hopkins University Melanoma Program director Suzanne Topalian said at one session.

Yet Mills noted the above-mentioned high mutation rates (generating startlingly good prognoses in some endometrial cancers) are creating such genetic chaos that they may be going “immunogenic,” arousing immune system attack. “This may be one of those rare cancers where the immune system plays a role,” he said.

Thus this cancer may require the skills of both geneticists and immunologists.

Indeed, so much crosstalk is needed, the NCI emailed thousands of grantees, asking for ideas to create “appropriate knowledge networks,” said Varmus.

The shared goal, he said, is nothing less than “creating a new taxonomy for cancer.”

(The TCGA was launched in 2009. It has published six major studies so far. One found that half of all squamous cell lung cancers possess mutations that could be treated with drugs that are already in the pipeline, or are easily synthesized. Another study found striking similarities between some breast and ovarian cancers, resulting in at least 40 new drug targets. A study on endometrial cancer will be published soon. See http://cancergenome.nih.gov/.)