Cancer patients are often told by doctors about how long they are likely to live and how well they respond to treatments. These, however, are tricky estimates. Now a new statistical method promises to provide more accurate information on how long a cancer patient may survive.
Researchers from UCLA have developed the new tool called Survival Analysis of mRNA Isoform Variation (SURVIV), which models the measurement uncertainty of mRNA isoform ratio in RNA-sequencing data in order to predict patients’ survival time.
Harnessing biomedical big data, the tool was tested on six cancer types: cancers of the breast, brain (highly aggressive and less aggressive), lungs, ovaries, and kidney. The data encompassing cancer’s molecular and clinical profiles led scientists to identify new biomarkers that surround cancer prognosis as well as treatment.
“[W]e found that isoform-based predictions work consistently better than the conventional gene-based predictions in predicting survival time," lead study author Dr. Yi Xing told Medscape, adding that it could take one to three years for their innovation to be used in clinical settings.
Spending over two years developing its algorithm, the team used samples from over 2,600 cancer patients and reported the identification of some 200 isoforms linked to survival time for patients of breast cancer. Some isoforms predict longer survival while others are tied to shorter times.
Using a metric known as C-index, the scientists evaluated how the survival predictors performed, discovering that their novel tool outperformed conventional gene-based predictions across simulation studies for the six cancer types.
These are surprising findings, said Xing, because contrary to what was previously thought, isoform ratios offer a stronger molecular signature of cancer than overall abundance of genes.
A human gene, he added, usually produced seven to 10 isoforms. In the case of cancer, a single gene sometimes creates two isoforms, each promoting or repressing metastasis or the spread of disease to other organs or parts of the body. To understand the differences between these two is deemed crucial in fighting the condition.
“We have just scratched the surface,” said Xing, mentioning their plan to apply their method to much bigger datasets across more types of cancer in order to develop more dependable isoform-based patient survival predictors.
The findings were detailed June 9 in the journal Nature Communications.
In related news, a separate study recently made good use of web searches to possibly detect cancer early. The team from Microsoft demonstrated that these online searches can be useful in predicting a future diagnosis for pancreatic cancer.
Photo: Peter Stevens | Flickr
