Abstract 2394: Survival and response modeling using diverse mouse mammary tumor models

Abstract Triple-negative breast cancer (TNBC) is a clinically heterogeneous subtype of breast cancer with poor outcomes and limited molecular predictors of survival. To better understand TNBC tumor growth dynamics and to identify biomarkers of treatment response, we developed a diverse panel of 26 mouse mammary tumor models (syngeneic transplants (n = 23) and autochthonous (n = 3)) that are immune-competent and reflect the molecular subtypes seen in human TNBC. These 26 tumor models exhibit distinct baseline tumor growth rates, drug sensitivities, and molecular profiles, which we have extensively characterized through bulk RNA-seq, scRNA-seq, DNA exome sequencing, and mass spectrometry-based kinase proteomics. This comprehensive, multi-platform data set provides a unique resource to perform machine learning analyses for response and survival predictions. Using this murine dataset, we first identified survival-associated gene expression features through Cox proportional hazards modeling. Murine tumors with high expression of basal-like subtype gene signatures were associated with shorter survival while luminal subtype and immune cell features correlated with longer survival; these findings closely align with patterns observed in human breast cancer. Second, we trained an Elastic Net-based survival predictor using untreated tumor RNA-seq data from each of the 26 mouse models and validated its prognostic utility using human breast cancer datasets such as SCAN-B (human set c-index = 0.63). Additionally, our mouse tumor models dataset includes efficacy testing for immune checkpoint inhibitors (ICI; anti-PD1/anti-CTLA4) and MEK inhibitor therapy (trametinib). Pre-treatment tumor RNA-seq analyses identified features associated with greater survival benefit in ICI-treated models, including high PDL1 mRNA and multiple T-cell and B-cell associated gene expression signatures. Similar analyses for MEK inhibition revealed key resistance-associated genes, such as Akt1, whose high baseline expression correlated with MEKi resistance. DNA copy number analyses of the murine tumor data identified distinct chromosomal regions linked to MEKi sensitivity (i.e. chr 12q) and resistance (i.e. chr 2q). These genomic alterations overlapped the gene expression associations, suggesting potential biomarkers for MEKi response. Ongoing work involves developing predictive Elastic Net models for both ICI and MEKi sensitivity, which will be validated using available human clinical trial datasets. Overall, this work demonstrates the translational potential of this diverse mouse models dataset as a platform for biomarker discovery and therapy optimization. Citation Format: Matthew D. Sutcliffe, Kevin R. Mott, Denis O. Okumu, Michael P. East, Gary L. Johnson, Shawn M. Gomez, Timothy C. Elston, Charles M. Perou. Survival and response modeling using diverse mouse mammary tumor models [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 2394.