Abstract 979: Development of a methylation-based classifier to identify pancreatic adenocarcinoma subtype

Background: Pancreatic ductal adenocarcinoma (PDAC) demonstrates variable prognosis and response to first-line FOLFIRINOX therapy depending on molecular subtype; patients with classical tumors showed a greater response compared to basal-like tumors. Identifying molecular subtype may inform treatment approaches, but current subtyping schemas rely on tumor biopsies which are invasive and can be technically challenging. Blood based epigenetic assays are commercially available, and identification of subtype-associated epigenetic modifications may provide a path toward blood-based PDAC molecular subtyping. We analyzed methylation levels and developed a subtype classifier using methylation data on PDAC tumors. Using this classifier, we then developed a methylation-specific qPCR assay to subtype PDAC tumors by analyzing methylation at particular sites in tumor DNA which may now be further adapted to more facile liquid biopsy approaches.

Methods: Using methylation data (n=150) from TCGA PAAD, we applied a random forest (RF) model followed by a k-Top Scoring Pairs (k-TSP) algorithm to develop a methylation site-based classifier that allows prediction of subtype by comparing methylation levels between paired sites. The classifier was validated on two independent datasets, including ICGC (n=82) and patient-derived xenografts (PDXs) (n=21). We designed methylation-specific qPCR primers for these sites and validated their specificity with fully methylated or fully unmethylated DNA. Using genomic DNA extracted from 10 PDAC PDXs of known molecular subtype, we obtained methylation levels at sites identified by the classifier via a SYBR Green-based qPCR assay. We then used our classifier to predict subtype of the PDX tumors.

Results: We validated our RF-kTSP methylation subtype classifier on two independent cohorts and predicted subtype using 20 methylation sites. In ICGC, we found balanced accuracy=0.94, AUROC=0.98, AU-PR=0.89. In PDX PDAC tumors, we found balanced accuracy=1, AUROC=1, AU-PR=1. We designed 14/20 primers for qPCR and confirmed specificity for methylated DNA. Using 14/20 methylation sites, we achieved 80% correct subtype prediction of 10 PDX tumors (5/5 basal-like, 3/5 classical).

Conclusion: We have developed a methylation-based classifier that accurately and replicably predicts PDAC subtype. We continue to develop a qPCR assay for PDAC subtype that so far has 80% accuracy using 14/20 probes. We are completing validation of the remaining 6 primers and correlating methylation levels at each site with those obtained from EPIC array methylation analysis. The long-term goal is to apply our complete set of primers in a digital droplet PCR assay that will be amenable for subtyping using circulating tumor DNA (ctDNA).

Citation Format: Zachary Schrank, Des Weighill, Hannah Thel, Hannah Trembath, Ashley Morrison, Jen Jen Yeh. Development of a methylation-based classifier to identify pancreatic adenocarcinoma subtype [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 979.

The Network Zoo: a multilingual package for the inference and analysis of gene regulatory networks

Inference and analysis of gene regulatory networks (GRNs) require software that integrates multi-omic data from various sources. The Network Zoo (netZoo; netzoo.github.io) is a collection of open-source methods to infer GRNs, conduct differential network analyses, estimate community structure, and explore the transitions between biological states. The netZoo builds on our ongoing development of network methods, harmonizing the implementations in various computing languages and between methods to allow better integration of these tools into analytical pipelines. We demonstrate the utility using multi-omic data from the Cancer Cell Line Encyclopedia. We will continue to expand the netZoo to incorporate additional methods.