Abstract B038: Spatial-Tx: Multi-modal computational framework for predicting spatial drug response from tumor tissues

Abstract Background: Understanding spatial heterogeneity within tumor microenvironments is crucial for accurately predicting therapeutic responses. Traditional transcriptomic methods often overlook essential spatial context, limiting the precision of personalized cancer treatments. Methods: We developed Spatial-Tx, a computational framework integrating three distinct modalities that include gene expression data, spatial coordinates, and histological imaging features using advanced graph transformer architectures and adversarial domain adaptation. Spatial-Tx leverages extensive pharmacogenomic data (GDSC and CCLE databases) for knowledge transfer, enabling accurate predictions of spatial drug responses in tumor tissues. Results: Spatial-Tx significantly enhanced the accuracy of spatially resolved drug response predictions compared to traditional single-modal approaches. Our method identified candidate spatial therapeutic niches and highlighted putative resistance-linked regions within the tumor tissues. Conclusion: Spatial-Tx provides a robust computational platform for precision oncology, offering the potential to tailor therapeutic strategies to the complex spatial architectures of tumors, thereby advancing personalized cancer treatments. Citation Format: Kayode Raheem, Sushil Kumar. Shakyawar, Patel Jai, Nagarajan Nagasundaran, Sajja Balasrinivasa, Chittibabu Guda. Spatial-Tx: Multi-modal computational framework for predicting spatial drug response from tumor tissues abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Cancer Evolution: The Dynamics of Progression and Persistence; 2025 Dec 4-6; Albuquerque, NM. Philadelphia (PA): AACR; Cancer Res 2025;85 (23Suppl): Abstract nr B038.