Abstract B111: Developing prognostic signatures (ep-Sigs) through epigenetic predictive markers in pancreatic ductal adenocarcinoma (PDA) – A pilot study based on The Cancer Genome Atlas (TCGA) data

Abstract

Abstract Current genomic testing (mutation profiling) has not proven beneficial for optimal treatment selection or for monitoring drug resistance, which is critical for neoadjuvant therapy (NAT) in borderline resectable/resectable PDA and first-line therapy for advanced PDA. Failure to administer an effective chemotherapy combination results in early progression of the disease, life-threatening complications, and a decline in the patient’s performance status, rendering them ineligible for further anti-cancer treatment. We hypothesized that our panel of epigenetic markers (DNA-methylation changes) is a surrogate for protein/gene expression and will develop an ep-Sigs that also has predictive value. The genes in this panel have been curated from extant literature and belong to one of two broad categories: one with known prognostic value (PvG) and the other with preclinical or clinical evidence of influencing one of the drugs (DrG) used in PDA management (gemcitabine (gem), 5-fluorouracil (5FU), oxaliplatin (Ox), irinotecan (Iri), nab-paclitaxel (NP), paclitaxel (Tax), and cisplatin (Cis)). Some genes/proteins were critical for more than one drug. To accomplish this, we tested the panel in the TCGA pancreatic adenocarcinoma (PAAD) data with 182 patients. For univariate survival analysis (UVA), we used Cox regression models for each marker. For multivariate survival analysis (MVA), we used penalized Cox regression models (i.e., LASSO, Elastic-Net) for identifying marker signatures. The area under the ROC Curve (AUC) was calculated to assess each signature’s predictive performance. We identified 301 cytosines followed by guanine residue (CpG) sites among the 67 genes (gem=35, 5FU=13, 5FU+gem=10, NP=6, Cis=6, and PvG=23) in the panel with significant (p<0.05) association with overall survival (OS) through UVA. We identified five ep-Sigs through MVA that could predict poor OS (p<0.05), which are a combination of markers from PvG and DrG. The number of markers and relevant DrG in them are as follows, 15 & Gem/Cis; 30 & Gem/5FU/Cis/Tax/Iri; 28 & Gem/5FU/NP/Cis/Tax/Iri; 23 & Gem/5FU; 29 & Gem/5FU/NP. We did a similar analysis for RNA-sequencing data from TCGA. We identified 17 markers and 9 signatures on UVA and MVA, respectively. Interestingly, the expression of 100/105 markers in the panel was significantly (p=4.19e-66) high in PDA compared to normal tissues, indicating this panel has diagnostic value too. We believe this project is the first step for this approach, and this concept needs to be evolved with prospective studies and larger samples. The ep-Sigs with a panel of epigenetic markers with a proven influence on resistance to frequently used drugs could help in developing meaningful models that change the approach to managing PDA, especially for NAT. Citation Format: Ashish Manne, Wancai Yang, Davendra Sohal, Ashwini Esnakula, Lianbo Yu. Developing prognostic signatures (ep-Sigs) through epigenetic predictive markers in pancreatic ductal adenocarcinoma (PDA) – A pilot study based on The Cancer Genome Atlas (TCGA) data [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl):Abstract nr B111.