Abstract 5924: A Tumor-specific, methylation-based algorithm to identify <i>MTAP</i> gene deletions via tissue-free circulating tumor DNA

Abstract Background: Methylthioadenosine phosphorylase (MTAP) is a tumor suppressor involved in the methionine salvage cycle. The MTAP gene is adjacent to the tumor suppressor CDKN2A on the chromosomal locus, 9p21.3. In ∼15% of all cancers, the locus 9p21.3 is deleted, and MTAP is co-deleted with homozygous loss of CDKN2A in &amp;gt;80% of cases. Clinical trials investigating PRMT5 and MAT2A inhibitors have demonstrated potential for inducing synthetic lethality in MTAP-deleted tumors. Thus, the characterization of MTAP deletion in solid tumors is necessary to identify eligible patients. However, robust detection of biallelic gene deletion in liquid biopsies requires high levels of tumor shedding. Here we show performance of a methylation-informed MTAP deletion classifier able to resolve copy loss down to 1% tumor fraction. Methods: We performed genomic and epigenomic profiling across samples from three distinct pan-cancer cohorts with the Guardant Infinity platform: 1. Discovery cohort of 14, 895 minimal-residual disease (MRD) plasma samples from patients after curative-intent treatment for early-stage tumors to identify tumor-specific methylation at the 9p21 locus 2. Discovery cohort of 23, 143 plasma samples from late-stage tumors to identify which candidate biomarkers were informative as to MTAP status, and 3. Independent validation cohort of ∼14, 000 plasma samples from late-stage tumors to confirm the association of candidate biomarkers to MTAP status. Quantification of tumor-specific DNA was done using an analytically validated methylation-based methodology, and we obtained synchronous mutation, copy-number, and genome-wide methylation data from all samples. Results: We identified 6 genes within 1MB of the MTAP locus (CDKN2A, CDKN2A-DT, CDKN2B-AS1, DMRTA1, MIR31HG, and HACD4) that demonstrated cancer-specific promoter hypermethylation in our MRD discovery cohort. In our late-stage cancer cohort, we observed that hypermethylation of all 6 candidate biomarkers was mutually exclusive with MTAP homozygous deletion relative to distal sites. A logistic regression model utilizing methylation of the MTAP proximal sites along with distal site hypermethylation was trained on a subcohort of 297 samples balanced for MTAP deletion status and tumor fraction. On the independent multicancer validation cohort, the AUC of the MTAP predictive model was 0.91. Incorporation of methylation signals increased MTAP homdel detection in liquid by 1.3-fold to several fold depending on sample tumor fraction. Conclusion: We have developed a novel method for improving the sensitivity of MTAP biallelic deletion detection in liquid biopsies using tumor-specific methylation. These findings offer biological insight into the relationship between localized deletion and methylation patterns and provide a framework to improve deletion calling in other genes with critical roles in tumor biology. Citation Format: Sean Gordon, Evan Warner, Geethika Yalamanchili, Jill Tsai, Scott Shell, Shile Zhang, Martina Lefterova, Justin Odegaard, Darya Chudova. A Tumor-specific, methylation-based algorithm to identify MTAP gene deletions via tissue-free circulating tumor DNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5924.