检测MTAP纯合缺失的多组学方法可实现从血液样本高灵敏度检测

Introduction: The methylthioadenosine phosphorylase ( MTAP ) gene encodes a tumor suppressor and component of the methionine salvage pathway. MTAP homozygous deletion (HomDel) can induce synthetic lethality when either PRMT5, MAT2A, or both are inhibited. Successful detection of MTAP HomDel can, therefore, offer novel options for targeted therapy. Unfortunately, tumor tissue is often unavailable for genomic profiling and detection of MTAP HomDel. While cfDNA testing is available for CGP, detection of MTAP HomDel in plasma cfDNA is challenging, as detection by conventional coverage-based methods is masked by non-tumor DNA coverage, especially in low-tumor fraction samples, thus limiting sensitivity in these settings. Methods: We developed a multi-omic approach that integrates genomic and tumor-specific methylation features to enhance detection sensitivity for MTAP HomDel in circulating tumor DNA (ctDNA). Tumor-specific methylation features were identified by analyzing methylation profiles from 100,089 plasma samples spanning diverse solid tumor types and used to train a classifier optimized to detect MTAP HomDel using the Guardant360 Liquid test. The resulting tumor-specific methylation classifier was then combined with genomic copy-number calling to maximize sensitivity and specificity for MTAP HomDel detection in plasma. Analytical accuracy was assessed in 87 matched cfDNA and tissue samples (NSCLC = 40%, pancreatic = 9%, melanoma = 9%, breast = 7%, other = 35%) from patients with advanced cancer. MTAP HomDel and single copy number loss were treated as distinct categories. Consecutive samples tested in routine clinical care were used to determine detection rates. Limit of blank (LoB) specificity was established using 120 cancer-free donor cfDNA samples (30 ng input), and the 95% Limit of Detection (LoD) was determined via in silico dilutions of 50 high confidence MTAP HomDel samples. Results: Our multi-omic methodology achieved a positive predictive value (PPV) of 100% and negative percent agreement (NPA) of 100% above LoD (90.9% and 96.6% overall) against a tissue-based reference. No false positives were observed among 120 cancer-free donors. Assessment of cfDNA samples from patients with advanced cancer demonstrated that multi-omic MTAP HomDel detection was over 2.4 fold as sensitive as genomic-only methods. Conclusion: Augmenting genomic information with tumor-specific methylation significantly enhances the detection of MTAP HomDel in plasma cfDNA, especially in low-tumor-fraction samples. Future work may leverage similar methylation coverage on the Guardant360 Tissue test to enhance MTAP HomDel detection in low input and low tumor purity tissue samples. This approach has potential to not only improve patient access to PRMT5- and MAT2A-targeted therapies but also to other therapies that are guided by gene deletions.