Artificial intelligence can ‘smell’ the DNA of tumors in the blood and has learned to recognise seven forms. This is indicated by the test, called Delphi, which identifies the fragmented form that distinguishes the DNA of tumor cells. The results were published in the journal Nature by the Johns Hopkins Kimmel Cancer Center group coordinated by Alessandro Leal and Jillian Phallen.
This is a proof of principle capable of paving the way for a more precise and non-invasive early diagnosis. The Delfi test (DNA evaluation of fragments for early interception) has in fact identified the tumor DNA in 73% of the blood samples of 208 patients with tumors, at different stages, of breast, colorectal, lung, ovaries, pancreas, gastric duct and bile. Also good results on blood samples from 215 healthy volunteers, where the false positives were 4. Artificial intelligence allows us to recognize the DNA fragments of tumors and in 75% was able to identify the tissues from which they derive the identified cells.
Unlike liquid biopsies that look for mutations in DNA sequences inside tumor cells, the new study studies the way in which the DNA is ‘packed’ into the cell nucleus, looking for DNA fragments in different parts of the genome.
“The nucleus of healthy cells packs the DNA like a well-organised suitcase, in which each region of the genome is carefully arranged in different compartments. The nucleus of those tumors is more like disorderly suitcases, where the various pieces are haphazardly placed”, comments Leal. “For various reasons the genome of tumors is packed in a disorganized fashion. This means that when cancer cells die, they release their DNA chaotically in the blood,” Phallen adds. If further studies confirm its efficacy, this test could be used to monitor and investigate the origin of a tumor.