Scientists developed CancerLocator tool to diagnose cancer non-evasively. The study implies the analysis of blood samples was indicating whether they contain tumor DNA and also in which tissue the tumor is located. The new approach was called CancerLocator, and it is bound to detect circulating cell-free DNA. Then, it uses the genome-wide DNA methylation profile to determine whether it comes from a tumor.
Researchers developed the CancerLocator tool to diagnose people with cancer
Jasmine Zhou from the University of California, Los Angeles together with her colleagues developed a study and published it in Genome Biology. Their new method was able to separate non-cancer from cancer samples, being better than the classification approaches which used forest and support vector machine. This type of non-invasive method of diagnosing cancer could help people find out an earlier diagnosis.
The sooner the cancer is found, the better the opportunity patients have to fight the disease. Zhou argued that she and her team built a computer-powered test which can detect cancer, by also naming the type of cancer after just analyzing a blood sample. The new technique is still in its developmental phase and needs further validation. However, scientists indicate that the potential benefits for patients are great.
This test was proved to outperform other similar non-invasive tests
Zhou and her colleagues have collected the Cancer Genome Atlas DNA methylation information and managed to build a database which contains methylation markers that appear to be common across several types of cancer, like lung cancer, liver cancer, kidney cancer, colon cancer and breast cancer. They have also created a set of methylation markers common to healthy tissues. To develop this CancerLocator tool, they have chosen CpG clusters able to differentiate types of tumor and healthy plasma.
For a particular sample of plasma, they created a methylation profile by implementing the use of whole-genome bisulfite sequencing which was later used as the input of CancerLocator to predict whether the sample contained a tumor DNA. The input was based only on the selected CpG clusters. Scientists have tested this study on both real and simulated data.
On simulated information, they revealed that their method had a Pearson’s correlation coefficient estimated at 0.975, between the presupposed and the real proportions of circulating tumor DNA. Moreover, they also compared the performance of their tool to random classification approaches of forest and support vector machine using simulated data. The result indicated that the CancerLocator surpassed the performance of other tests.
Image source: wikipedia