ROSIE-Enabled Spatial Mapping Reveals Architectural Fragmentation and Immune Reprogramming in Lung Adenocarcinoma Evolution

Abstract The progression of lung adenocarcinoma (LUAD) from precancerous lesions to invasive carcinoma entails extensive remodeling of tissue architecture and immune microenvironments, driven by mechanisms that remain incompletely understood. To investigate these dynamics, we applied ROSIE (RObust in Silico Immunofluorescence), a deep learning framework that computationally infers multiplex biomarker expression directly from routine H&E-stained tissue slides. This approach enables scalable, high-resolution spatial profiling without the need for specialized staining or equipment. Using ROSIE, we analyzed a cohort comprising 114 human LUAD samples to reconstruct the evolving immune landscape during tumorigenesis. Our analysis revealed a fundamental immunological shift: adaptive immunity progressively recedes, giving way to a dominant innate and immunosuppressive state as lesions advance. Concurrently, we observed progressive architectural fragmentation that marks the transition to malignancy. These findings highlight the power of computational pathology to map cancer evolution at scale and pinpoint critical windows for therapeutic interception, laying the groundwork for more precise and proactive strategies in oncology. Competing Interest Statement The authors have declared no competing interest. Footnotes The manuscript has been updated to include pathway analysis, providing deeper biological insights.