Researchers are calling for adoption of a molecular-based classification system of cancer rather than the current tissue-of-origin methods, according to a study published Aug. 14 in Cell
. Utilization of the new molecular taxonomy would reclassify one in 10 cancer patients, which the authors say could lead to different treatment choices. “The refined molecular taxonomy we describe builds on centuries of pathology and genetic research,” write the authors, led by Katherine Hoadley, Ph.D., and colleagues from the Cancer Genome Atlas research network. “This initial Pan-Cancer-12 analysis lays the groundwork for a richer classification of tumors into molecularly defined subtypes unlike all prior cancer classification systems.” The researchers utilized data from six different “omics” platforms to look for molecular alterations shared across cancers arising from different tissues. This integrative analysis was completed on 3,527 samples from 12 tumor types referred to as the “Pan-Cancer-12” set. Cases were assayed by at least four of the six possible methods: whole-exome sequencing, DNA copy number, DNA methylation, mRNA expression, microRNA expression, and protein expression. Statistical analyses of the molecular data (both individually from each platform and integrated cross-platform) divided the tumors into clusters. Eleven integrated cancer subtypes were identified through cluster-of-cluster assignments (COCA). While five of these subtypes were consistent with tissue-of-origin classifications, several newly identified subtypes were seen across tissues, and some tissue-of-origin categories were split into multiple different molecular subtypes. Importantly, approximately 10 percent of cases were reclassified by the molecular taxonomy, with the COCA subtypes providing important molecular and tumor biology information that is significantly associated with prediction of clinical outcomes beyond tumor stage and primary tissue of origin. “We’re just appreciating the tip of the iceberg when considering the potential of this multi-platform type of genomic analysis,” co-senior author Christopher Benz, M.D., from the Buck Institute for Research on Aging (Novato, Calif.), said in a statement. “It could be that as many as 30 or 50 percent of cancers need to be reclassified” when more tumor samples and 20 tumor types are included in the next round of analysis. Benz believes that even when looking at the 12 current tumor types, the 10 percent reclassification rate in the current study is likely an underestimate due to the unequal representation of different tumors. “If our study had included as many bladder cancers as breast cancers, for example, we would have reclassified 30 percent,” Benz said. Although there were only 120 bladder cancer samples included, it proved to be the most diverse tumor type, with samples clustering into seven of the 11 COCA subtypes. Among the most dominant bladder subtypes, one was remarkably similar to lung adenocarcinomas and another was similar to head and neck squamous-cell cancers. Survival differences were dependent on subtype classification. Other specific findings included:
- The two most commonly mutated genes in the overall data set were TP53 (41 percent) and PIK3CA (20 percent), which were both prognostic, even across different tumor types.
- Regardless of molecular platform, cancers of nonepithelial origin appear most different from epithelial tumors. This confirms previously established differences between breast cancer subtypes (basal-like and luminal). In this study, the breast basal-like subtype exhibited similarities between lung squamous cell carcinoma and serous ovarian cancers.
While further validation of this taxonomy is needed, the authors are hopeful that the results will further propel clinical trial design to rely on genomic classification of tumors for eligibility. Takeaway: Classification of cancers based on a new integrated taxonomy, rather than tissue of origin, reveals common molecular features across tissues of origin and may be useful in reclassifying patients for treatment purposes.