Next, in collaboration with Prof. Guy Ron from the Racah Institute of Physics at the Hebrew University of Jerusalem, the scientists combined what they had revealed about the molecular biology of cancer with artificial intelligence algorithms, applying machine learning to the large data sets obtained from the two groups. The analysis was performed not only on all these cancer markers but on combinations of and relationships between them as well. To make sure their findings are not limited to colorectal cancer, the scientists also applied their technology to compare blood nucleosomes of healthy volunteers with those of 10 patients with pancreatic cancer.
“Our algorithm could tell the difference between the healthy and the patient groups at a record level of certainty for studies of this type – with 92 percent precision,” Shema says. The scientists call the new technology EPINUC, an acronym for “epigenetics of plasma-isolated nucleosomes.”
If supported by studies involving a greater number of patients, these findings could lead to a multiparameter blood test for detecting and diagnosing cancer using less than 1 ml of blood. In the future, because of the level of detail revealed in the analysis, the results of this blood test might also advance personalized medicine by suggesting the best treatments for each individual patient.
Shema sums up: “We’ve achieved a successful proof of concept for our method, which now needs to be confirmed in clinical trials. In the future, our multiparameter approach may serve to diagnose not only various cancers but also additional diseases that leave traces in the blood, such as autoimmune disorders or heart disease.”