Researchers across the Weill Cornell Medicine, NewYork-Presbyterian, the New York Genome Center and Memorial Sloan Kettering Cancer Center, USA, concluded in a new, compelling study that an AI-powered method for detecting circulating tumour DNA, has revealed greater sensitivity than standard methods.
Minimal residual disease refers to a select quantity of cancerous cells that remain in a patient during or after treatment. Improving detection tools is vital in fighting minimal residual disease, and ultimately improving early prevention of this disease.
With this in mind, Adam Widman, Memorial Sloan Kettering Center, New York, USA, and colleagues trialled a new machine-learning approach with the aim of improving the early detection rates of various cancers; including breast, lung, colorectal and myeloma among others. The model in question, is called MRD-EDGE. It is a machine-learning guided approach to whole-genome sequencing (WGS) that was trained to recognise patient-specific tumour mutations in circulating tumour DNA (ctDNA).
In one test, MRD-EDGE was tested on blood tests from 15 patients with colorectal cancer. After the patients’ surgery and chemotherapy, the system predicted residual cancer in nine individuals based on blood data. Months later, less sensitive methods confirmed cancer recurrence in five of these patients. Importantly, there were no false negatives; none of the patients identified as tumour DNA-free by MRD-EDGE experienced recurrence during the study period.
MRD-EDGE demonstrated comparable sensitivity in early-stage lung cancer and triple-negative breast cancer studies, detecting all but one recurrence early and monitoring tumour status during treatment. The researchers also showed that MRD-EDGE could detect mutant DNA in precancerous colorectal adenomas, from which colorectal tumours develop. Furthermore, MRD-EDGE could detect responses to immunotherapy in melanoma and lung cancer patients without pre-training on tumour sequencing data, weeks before standard X-ray imaging methods.
Helena Bradbury, EMJ
Reference
Widman et al. Ultrasensitive plasma-based monitoring of tumor burden using machine-learning-guided signal enrichment. Nature Medicine. 2024;30:1655-66.
