Using data from The Cancer Genome Altlas' comprehensive analysis of breast cancer, researchers from the University of Ferrara and Ohio State University have developed a 37-marker combined microRNA and mRNA signature that shows promise as a prognostic tool for patients with early disease.
The study's authors — the University of Ferrara's Steafano Volinia and OSU's Carlo Croce — published their results in the Proceedings of the National Academy of Sciences this week, describing how their team developed and tested the signature based on miRNA, mRNA, and DNA methylation sequencing data from TCGA's breast cancer analysis.
According to the researchers, the signature had an area under the receiver-operator characteristic of 0.74 at 60 months of overall survival in the TCGA cohort, with the highest prognostic power — an AUC of 0.77 — in early stage I and stage II tumors.
"The early stages are the most interesting because there is more and more early detection [in breast cancer] and so it's an important question whether you can really predict whether an early tumor progresses or not," Volinia, the study's first author, told GSN this week.
According to Volinia, the group initially was expecting that they would use the TCGA data to measure whether a few miRNAs — identified in earlier research as differentiated in in situ versus invasive ductal carcinomas — were really important for prognosis or not.
But working with the TCGA results, the researchers saw that they could combine miRNA and mRNA together to create a potentially more powerful prognostic predictor — the first such predictor, according to the authors, to include both types of RNA.
"We are definitely in debt to TCGA because this would have been impossible without that work," Volinia said.
In the study, Volinia and Croce used sequencing data from TCGA's analysis, focusing on those patients with invasive ductal carcinoma, to develop a signature of seven miRNAs — miR-103, miR-1307, miR-148b, miR-328, miR-484, miR-874, miR-93 — and 30 mRNAs that could predict patient outcomes, measured by overall survival.
In the full cohort of 466 subjects, the signature had an AUC of 0.74, while in 348 of the subjects who had stage I or stage II IDC tumors, it did even better, with an AUC of 0.77.
To validate the findings, Volinia and Croce then went on to measure the performance of the signature in other retrospective cohorts.
"[The] signature was very successful for this [TCGA] cohort," Volinia said. "But each lab that develops a signature in a cohort can do a very good job within that same cohort, so we wanted to validate the signature in other cohorts."
Overall, the researchers tested the signature in eight cohorts with more than 2,000 additional samples in total. One validation cohort — of 207 breast cancer patients — had both mRNA and miRNA data available, though only for nine miRNAs and 11 mRNAs. Though this did not allow the team to use the full signature, the predictor still showed a significant AUC of 0.65, according to the study authors.
In the rest of the cohorts only mRNA data was available, but the group found that even using only the mRNA portion of the prognostic signature was "significantly predictive for outcome."
Finally, Volinia said, the team compared its predictor to the performance of gene sets gleaned from a number of other prognostic panels used for risk stratification in breast cancer, including Oncotype DX and MammaPrint.
According to the authors, the miRNA/mRNA signature gleaned from the TCGA data did significantly better than many of these in the TCGA cohort itself, while several others outperformed it or at least matched it in some of the validation cohorts.
Volinia said that from the TCGA data, it appears that the miRNA component of the group's signature is playing an important role, but because the team was not able to validate the miRNAs in most of the validation cohorts, it is hard to know that for sure yet.
"In the TCGA set, we separated [the signature] into its microRNA and mRNA components, and though they both had strength, together they behaved better," he said.
"For sure, the miRNA component adds to the signature. Unfortunately we didn’t have miRNA from the other cohorts, so [we can't really tell] how important the miRNA component is."
According to Volinia, the initial plan for the research was modest, so there wasn't much thought about commercial potential of the signature. But because of the promising results, the team has now applied for patents on the signature.
"If we can confirm [our results] with more work, it could be amenable [to commercial use] if there is interest in it from the market," he said. "But because there are so many competitors it may not even be feasible. It depends on interest from the companies."
Now, Volinia said he is working on testing methods for simultaneous measurement of miRNA and mRNA, and planning to study whether any of these techniques can replicate the results from testing using separate assays.
He said his group is also looking into the possibility of early detection of the miRNA signature from blood.