ML-based approach to identify cancers from repeat elements of genetic code

The ‘repeats’ of DNA sequences found in chromosomes, aptly called ‘junk DNA’ or ‘dark matter’ and feared for their cancer-causing attributes, have long posed a stiff challenge for researchers: how to identify and characterize them. Johns Hopkins Kimmel Cancer Center researchers have now developed a Machine Learning-based approach to identify these elements both in cancerous tissue, and in cell-free DNA (cfDNA)—fragments shed from tumors and floating in the bloodstream.

Going forward, this new method could provide a non-invasive means of detecting cancers or monitoring response to therapy. Throwing light on the hitherto ‘dark genome’, it has provided a proof-of-concept for using genome-wide repeat landscapes as tissue and blood-based biomarkers for cancer detection, characterization and monitoring.