Precision cancer treatments have improved patient survival rates by targeting underlying molecular drivers of disease, but only for those with commonly known mutations like BRCA1/2. Now researchers at Harvard Medical School have devised an algorithm to target such treatments to patients without these telltale mutations. The algorithm examines patterns of mutations across multiple genes to infer underlying molecular drivers, identifying a wider selection of patients likely to respond to current precision treatments. Their work, reported in Nature Genetics on April 15th, promises to expand the reach of precision medicine across various cancer types.
Different environmental and molecular drivers of cancer generate specific patterns of mutation across the genome. For example, ultraviolet light tends to cause C-to-T base pair mutations, while smoking causes other genomic changes. Meanwhile, internal molecular drivers, like BRCA1/2 malfunction, also generate unique mutational signatures.
Altogether, there are nearly 40 distinct mutational signatures, which have been characterized through whole genome and exome sequencing of thousands of tumor samples. While the underlying drivers of most of these signatures is yet unknown, some have been linked to specific cancer mutations and molecular drivers. These signatures give researchers clues to what is driving the disease, and how to thwart it. “Mutational signature analysis has emerged as a powerful approach for investigating the processes that generate somatic (cancer) mutations,” says said study senior author Peter Park, professor of biomedical informatics in the Blavatnik Institute at HMS.
Specifically, BRCA1/2 mutations are associated with a particular signature which indicates underlying defects in homologous-repair (HR) machinery, a type of DNA repair mechanism. Tumors with this type of mutation respond well to PARP-inhibitors, which target this defect. But this signatures is also present in some BRCA negative cell lines and patient tumors, suggesting that defects in HR machinery may underlie a wider, but unknown, selection of patients. “This is clinically relevant because those without a mutation in a known HR gene but that present Signature 3 may benefit from treatments that target selective vulnerabilities of HR-deficient cancers.” says Park. “A recent study using breast cancer organoids for example has shown that a high burden of signature 3 mutations is associated with a better response to PARP inhibitors.”
Park and his team develop an advanced algorithm to identify such patients based solely on data provided by standard genetic testing panels—a significant advance geared towards clinical utility. “For the most common genetic testing platform in oncology clinics- targeted sequencing panels- the number of mutations identifiable is far too small for standard signature analysis. Whereas previous methods required whole-genome or exome data, our method detects the HR-deficiency signature even from low mutation counts, using a likelihood-based measure combined with machine learning techniques,” Park explains.
The new technique enables the team to identify far more patients with underlying HR-deficiency, through standard clinical methods. The team estimates that for every patient identified with a BRCA1/2 mutation, there may be twice as many with undetected HR-defects. Considering breast cancer alone, that amounts to 27,000 – 54,000 patients per year, based on 2018 data. The signature is likely even more widespread in ovarian cancer patients. Many other cancers, including pancreatic cancer, are known to be fueled by HR-deficiencies, but the prevalence in the population remain unknown.
“By enabling panel-based identification of mutational signatures, our method substantially increases the number of patients that may be considered for treatments targeting HR deficiency,” says Park. Indeed, the team finds that cell lines and ovarian cancer patients deemed HR-deficient by the algorithm responded better to PARP-inhibitors and survived longer, confirming their hypothesis.
“We have spoken with many clinicians in the past months and we have started multiple collaborations in which additional patients in clinical trials will be given the drug based on our predictions,” Park said. “We think we could make a real impact in cancer care with this computational method.”