Lessons of TCGA
Philosopher Ponders Lessons, Unknowns of Mapping Cancer Genome
Delving into the genetics of cancer can seem an endless challenge. Despite all its complexity, there’s been great progress thanks to advances in sequencing technologies. Over the past decade, cancer genome sequencing studies have uncovered many new cancer genes, pathways and targets. Still, much remains unknown. This begs the question: Can there ever really be a complete cancer genome?
Dr. Anya Plutynski is exploring this conundrum as a follow-up to her new book Explaining Cancer, which probes the trajectory of cancer, classification, risk and what constitutes progress in cancer research. Plutynski, a philosophy professor and director of graduate studies in philosophy at Washington University in St. Louis, discussed her new research project at a recent NHGRI History of Genomics and Molecular Biology lecture.
“Cancer genomics is discovery science,” said Plutynski, “confirming and deepening our understanding of what we already knew, but also raising many questions, suggesting novel hypotheses and forcing refining and reframing of goals.”
A “complete” cancer genome doesn’t, and cannot, exist, posits Plutynski. “Cancer genomes are samples of dynamic, heterogeneous classes of disease processes,” she said. Can sequencing account for the massive, potentially infinite number of variations and mutations of the disease? And, from a practical standpoint, what sample size would be truly representative?
Plutynski didn’t have to go far to find experts on this topic. She recently interviewed cancer genomics researchers at Washington University School of Medicine in St. Louis to learn more about the outcomes of The Cancer Genome Atlas (TCGA). Dr. Timothy Ley, who studies acute myeloid leukemia, and Dr. Ramaswamy Govindan, who specializes in lung cancer, were researchers with TCGA, an NCI-NHGRI initiative launched in 2006 that mapped genomes of lung, brain and ovarian cancers. Participating cancer centers across the country ultimately characterized the genomics of more than 11,000 tumors from 33 cancer types.
Over the course of TCGA, which concluded this year, researchers benefited from increased sequencing speed and decreasing costs of sequencing the whole genome. As time went on, researchers learned even more with many more patients recruited and better biopsies and analytic tools. One heralded accomplishment was a more comprehensive molecular understanding of breast cancer.
At the same time, there was confusion over how to compare and interpret the data in this new, quickly changing field.
At first, researchers had to learn how to rapidly align and analyze whole genome data and how to compare a tumor with a matched normal genome, Ley told Plutynski. Ley led a team at Washington University’s Genome Institute that sequenced the first cancer genome using massively parallel sequencing. Those results were published in 2008 as a blueprint, though admittedly, he wasn’t sure how to interpret what his team had found without a lot more data from other patients.
As technology advanced and new knowledge emerged, open-ended goals and targets continued to change. Some researchers pursued larger sample sizes. Govindan argued that more samples didn’t necessarily yield more information. Well-designed clinical studies with appropriately collected samples are more important than just a random collection of tumor specimens, he said.
“There was an expectation that larger sample sizes increased the power to detect mutations,” said Plutynski. “Subsequently, [researchers] tried to refine their analyses, taking heterogeneity into account.”
Before and after TCGA, researchers have had to manage their expectations. Ley said he thought when all the mutations were assembled, they could better assess risk and better decide how to treat each patient. But it didn’t work out that way. Clonal heterogeneity complicated things.
The mutation rate per cancer, and heterogeneity in mutation processes across and within cancers, was an ongoing challenge that affected analysis, including the rates of false positives, said Plutynski. Researchers also had to account for the mounds of extra-genomic information; tumors don’t exist in a vacuum but are surrounded by, and interact with, all kinds of cells and fluids in the body.
“They found more complexity than expected,” said Plutynski. But thanks to technological advances and improved data-sharing, she said, “researchers are starting to understand how everything interacts over time.”
In discussions with Ley and Govindan, Plutynski learned of some other lessons and challenges confronted by TCGA researchers. It’s important to look not only at patients with newly diagnosed tumors, but also at resistant and relapsed tumors, they said. Researchers also must address the constraint of taking tumor samples at one point in time.
“If we’re taking samples for targeted genes,” she asked them, “are you also looking at these genes after chemotherapy?” That’s happening now and it holds great promise, they said.
The end of TCGA was really a beginning, a framework to inspire further discussion about cancer, changing treatment options and future genomic studies, said Plutynski. TCGA data provided a critical reference point, Govindan told Plutynski, so that researchers could know what new questions to ask going forward.