NEW YORK (GenomeWeb News) – Drawing on Cancer Genome Atlas Research data, the University of Texas MD Anderson Cancer Center's Roel Verhaak and an international team of researchers expanded on subtype and survival gene expression signatures for patients with high-grade serous ovarian carcinoma.
As they reported online today in The Journal of Clinical Investigation, the signatures, when combined, make up a model of high-grade serous ovarian carcinoma prognosis the researchers dubbed CLOVAR, for Classification of Ovarian Cancer.
Such a tool, they added, could help researchers better understand tumorigenesis of ovarian cancers and eventually improve patient outcomes.
"Studying HGS-OvCa in light of expression subtypes exposes features of ovarian carcinoma pathology that would otherwise have been neglected," Verhaak and his colleagues wrote. "The comprehensive CLOVAR framework presented here should provide a basis for improved understanding of ovarian carcinoma tumorigenesis that may ultimately lead to more effective treatments."
To develop CLOVAR, the researchers built off of the survival gene expression signature and the four subtype gene expression signatures the Cancer Genome Atlas Research network reported in Nature last year. Those subtype signatures divvied the high-grade serous ovarian carcinoma into immunoreactive, differentiated, proliferative, or mesenchymal categories. In this study, the researchers expanded those signatures and integrated them into CLOVAR.
Further, tumor samples, Verhaak and his colleagues reported, don't always fall neatly into one of the categories as they can have characteristics of different profiles.
"The overlap in gene signature scores suggests that HGS-OvCa does not consist of mutually exclusive expression subtypes, but that each tumor sample is represented by multiple signatures at different levels of activation," the researchers wrote. "This pattern may reflect a higher level of homogeneity than is seen in other tumor types, such as glioblastoma and breast cancer."
Rather than relying on the 193 genes previously used in the signature of survival, the researchers homed in on a smaller set, choosing the 100 genes whose expression were the most — or least — indicative of good prognosis from the full 481 TCGA high-grade serous ovarian carcinoma sample set. A signature of this size, they added, would allow it to be implemented on a medium-throughput gene expression quantification system.
Applied to a validation set, this new survival signature could stratify high-grade serous ovarian carcinoma samples into a good prognosis group and a poor prognosis group. The difference in survival between those groups, the researchers added, was statistically significant.
Similarly, the researchers then revamped the subtype gene expression signature by whittling the initial list of 800 genes down to 100 genes. That signature, the investigators noted, could classify the TCGA cohort with a cross-validation error rate of 8 percent.
To roll those gene expression signatures together to classify patients, the investigators developed a multiple covariate model based on the TCGA dataset and used the CLOVAR survival, immunoreactive, and mesenchymal gene signature scores as the variables. They then tested that model, along with other prognostic factors such as BRCA status, disease stage, or age to try to improve the model's predictions. Their results suggested that combining the CLOVAR survival, immunoreactive, and mesenchymal scores with BRCA1/BRCA2 mutation status "provides optimal outcome predictions."
Further, the association they uncovered between the CLOVAR survival, immunoreactive, and mesenchymal gene signature scores "suggests an active role for the stromal tumor microenvironment in the pathogenesis of HGS-OvCa" and may indicate possible targets for cancer therapies.
Additionally, the researchers noted that the group with the worst outcome — patients with tumors classified using the CLOVAR tools as both mesenchymal and with a poor prognosis — had a median survival time of 23 months and a 63 percent platinum resistance rate. Other cases had a median survival time of 46 months and a 23 percent platinum resistance rate.
This prognostic model must still be validated in a prospective study, the researchers cautioned. "A prospective study would be most revealing when assessing the predictive capacities of the CLOVAR signatures in conjunction with other prognostic factors, such as BRCA mutation status, age, grade, and residual disease," they added.