Multiplex Methylation Detection Assays Using a Blocking FRET Probe with Machine Learning-Assisted Quantitative Melting Curve Method Targeting Early-Stage Breast Cancer

Breast cancer (BC) is the second most prevalent form of cancer, and poses a significant threat to public health. DNA methylation is an ideal marker for the early detection of BC. Fluorescence quantitative polymerase chain reaction (PCR)-based DNA methylation detection is simpler and faster but is constrained by its multiplexing capability and specificity. To address this, we developed a multiplex quantitative methylation PCR assay for the simultaneous analysis of methylation status at multiple sites specific to BC (cg11754974, cg13828440, cg18637238, and cg16652347). The machine learning model was trained using 1200 cases of multipeak data to enhance the melting curve resolution. Performance testing demonstrated the method's ability to selectively amplify methylated genes at a DNA concentration of 1 x 10(5) copies mu L-1, with high replicability (coefficient of variation <5 %) and mutation detection capabilities as low as 10 %. When applied to 80 clinical BC samples, the assay effectively distinguished patients with early-stage BC from normal controls, achieving an area under the curve of 0.8938, sensitivity of 83.33 %, and specificity of 84.62 %. Our essay exhibits superior clinical performance when compared to the quantitative methylation-specific PCR assays for noninvasive detection of early-stage BC, which is poised to become a favorable clinical diagnostic method for early-stage BC owing to its simplicity, speed, and capacity to improve diagnostic accuracy.