Abstract A103: Development of the cellular therapy barrier framework (CTBF): A computational transcriptomic approach to stratify pancreatic adenocarcinoma for precision cellular immunotherapy
Khaoula MazouziAbstract
Introduction:
Pancreatic ductal adenocarcinoma (PDAC) remains one of the most treatment-resistant malignancies, with limited efficacy of cellular immunotherapy due to multiple interacting barriers within the tumor microenvironment. Existing molecular classifications describe tumor biology but do not comprehensively characterize resistance mechanisms relevant to cellular immunotherapy. We developed the Cellular Therapy Barrier Framework (CTBF) to identify biologically distinct resistance phenotypes and support precision cellular immunotherapy.
Methods:
RNA-seq data from 183 TCGA-PAAD tumors were obtained through the Genomic Data Commons. CTBF integrates seven biologically curated transcriptomic modules representing stromal exclusion, myeloid suppression, T/NK-cell exhaustion, immune exclusion, hypoxia/metabolic stress, immunosuppressive cytokines, and target availability. Following normalization, single-sample Gene Set Enrichment Analysis (ssGSEA) was performed using GSVA, and Z-score normalization generated a standardized 183 × 7 patient-by-barrier matrix. Each tumor was represented in a seven-dimensional barrier space and clustered using unsupervised methods. Cluster stability and the optimal number of phenotypes were assessed using the elbow method, average silhouette width, gap statistic, K-means clustering, and consensus clustering. Additional analyses included principal component analysis (PCA), pairwise barrier correlation, and gene overlap assessment using the Jaccard similarity index. Generative artificial intelligence (ChatGPT, OpenAI) assisted in R code development, and Grammarly was used for proofreading. All code, analyses, and text were reviewed and approved by the authors prior to submission.
Results:
CTBF successfully quantified all seven resistance modules across 183 PDAC tumors, achieving 100% gene coverage for all curated signatures. Correlation analysis identified a coordinated immunosuppressive network, with strong positive associations among stromal exclusion, myeloid suppression, T/NK-cell exhaustion, and immunosuppressive cytokines, whereas target availability showed inverse relationships with several immunosuppressive barriers. Jaccard similarity analysis demonstrated minimal overlap among gene signatures, indicating that each module captured a distinct biological process. PCA demonstrated clear organization of tumors within the seven-dimensional barrier space. Multiple complementary clustering approaches consistently supported a three-phenotype solution, identifying three reproducible Cellular Therapy Barrier Phenotypes (CTBF-P1, P2, P3) representing distinct cellular therapy resistance landscapes.
Conclusion:
CTBF establishes a novel computational framework for biologically informed stratification of PDAC according to mechanisms of resistance to cellular immunotherapy. By integrating multiple resistance pathways into a unified transcriptomic model, CTBF provides a scalable foundation for precision cellular immunotherapy and future development of personalized therapeutic decision-support tools.
Citation Format:
Khaoula Mazouzi. Development of the cellular therapy barrier framework (CTBF): A computational transcriptomic approach to stratify pancreatic adenocarcinoma for precision cellular immunotherapy [abstract]. In: Proceedings of the AACR Conference on Pancreatic Cancer: New Frontiers in Biology and Therapeutic Development; 2026 Sep 25-28; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(18_Suppl_2):Abstract nr A103.