Cell-free DNA (cfDNA), predominantly derived from apoptotic or necrotic cells, is emerging as a potential biomarker due to non-invasiveness and cost-effectiveness1. Additionally, another type of cfDNA such as cell-free mitochondrial DNA (cf-mtDNA) that is released from cells under stress or damage condition, may play a key role in the immune system. Thus, the presence of cfDNA in the circulation of cancer patients with differential observations between the cancer and the healthy individuals may be used to improve diagnosis by liquid biopsies2. In the past decades, liquid biopsy has been significantly advanced, particularly in detecting cfDNA, which has opened a new research era in cancer diagnostics worldwide. However, there is a big hindrance in integrating cfDNA analysis with clinical diagnosis, For example, a high rate of false positives is caused by insensitive assays, which may be improved by refining the processing of sequencing data. That is due to the technical differences of feature analysis (e.g., fragmentomics and methylation patterns) and their heterogeneity limited the application of cfDNA in the clinics3–5. Currently, there are many efforts on the way to solve the problems. Especially, the top priority is to enhance assay sensitivity and specificity and to adapt analytical techniques in oncology.Nowadays, liquid biopsy technology based on circulating free DNA (cfDNA) brings new hope for future clinical applications such as minimally invasive early screening for tumors, with the key to effectiveness lying in the integration of next-generation sequencing (NGS) with advanced separation technologies. A comparison between the RNA expression datasets and the cfDNA fragment frequencies has revealed distinct genomic characteristics between cancer patients and healthy individuals. Research has found that there are specific fragmentation patterns of cfDNA in the promoter regions of some highly expressed genes6,7. Notably, cfSort is the first supervised tissue deconvolution approach powered by deep learning, leveraging a high-resolution methylation atlas from 521 non-cancer tissue samples across 29 major human tissue types. This model accurately identifies the origin of cfDNA, demonstrating the clinical value of sequencing data in disease diagnosis and treatment monitoring8. In this perspective, we outline the current trend of cfDNA research and highlight its potential limitations in oncology (Fig. 1). It provides valuable insights for cancer diagnosis and management which may help pave the way for the future adoption of cfDNA in oncology.
Read full abstract