Abstract
BACKGROUND: An effective urine-based method for the diagnosis, differential diagnosis and prognosis of multiple myeloma (MM) has not yet been developed. Urine cell-free DNA (cfDNA) carrying cancer-specific genetic and epigenetic aberrations may enable a noninvasive "liquid biopsy" for diagnosis and monitoring of cancer. METHODS: We first identified MM-specific hydroxymethylcytosine signatures by comparing 64 MM patients, 23 amyloidosis (AM) patients and 59 healthy cohort. Then, we applied a machine learning algorithm to develop diagnostic and differential diagnosis model. Finally, the prognosis of MM patients was predicted based on their survival time at the last follow-up. RESULTS: We identified 11 5hmC markers using logistic regression algorithm could effectively diagnosis MM (AUC = 0.902), and achieved 85.00% specificity and 85.71% sensitivity. These 11 markers could also effectively differential diagnosis MM (AUC = 0.805) with 88.89% specificity and 73.08% sensitivity. In addition, the prognostic prediction model also effectively predicted the prognosis of patients with MM (p < 0.01), of which 4 differential markers (RAPGEF2, BRD1, TET2, TRAF3IP2) could independently predict the prognosis of MM. CONCLUSIONS: Together, our findings showed the value of urine cfDNA hydroxymethylcytosine markers in the diagnosis, differential diagnosis and prognosis of MM. Meantime, our study provides a promising and completely non-invasive method for the diagnosis, differential diagnosis and prognosis prediction of MM.