Abstract IA011: Decoding actionable tumor ecosystems through spatial multi-omics and AI

Abstract

Pancreatic ductal adenocarcinoma (PDAC) evolves within highly organized and dynamic tumor ecosystems in which malignant, stromal, and immune cell populations continuously interact in space. We conceptualize this spatial organization as a “cellular language,” where the identity, state, and spatial arrangement of individual cells collectively define the functional architecture of the tumor. Decoding this language may reveal how PDAC progresses, metastasizes, evades immunity, and responds to or resists therapy. In this presentation, I will discuss computational frameworks that integrate digital pathology, spatial multi-omics, multiplexed imaging, and artificial intelligence to resolve PDAC ecosystems across multiple biological scales. These approaches connect tumor-cell states and lineage plasticity with the organization of cancer-associated fibroblasts, immune populations, and multicellular neighborhoods, enabling us to define recurrent spatial niches and ecosystem states during tumor progression and metastatic evolution. Longitudinal and treatment-associated analyses further reveal how these ecosystems are remodeled by therapy and identify spatial features associated with therapeutic response and resistance. By integrating gigapixel histopathology with high-dimensional molecular and cellular measurements, AI provides an opportunity to move beyond descriptive spatial maps toward scalable models that recognize clinically relevant ecosystem states in routine tissue specimens. Together, these studies illustrate how moving beyond descriptive spatial maps to decode the functional grammar of PDAC ecosystems can uncover mechanisms of progression and therapeutic resistance, identify actionable multicellular interactions, and enable spatially informed precision oncology. The concepts and computational strategies presented here provide a broadly applicable framework for moving from high-dimensional spatial data to biologically interpretable and clinically actionable hypotheses, with relevance to studies of pancreatic cancer and other solid tumors.