Abstract 2757: Scale meets Speed: A transcriptome-wide spatial analysis that runs a rapid and interactive hypothesis-driven workflow with AtoMx® SIP.

Abstract The potential of whole-transcriptome spatial analysis to answer complex biological questions is often hindered by the sheer weight of its computational and analytical demands. As the number of cells, samples, and plex increases these computational challenges are compounded resulting in bottlenecks that slow the pace of discovery. Moreover, these complex datasets often require specialist data analysis expertise that can further slow “data-to-insight”. To bridge this gap, we created a new analysis paradigm that enables biologists and researchers with less analytical expertise to interact with their data directly, performing deep hypothesis-driven analysis in a rapid, conversational and accessible workflow. In this presentation we discuss three advancements arriving in the reimagined AtoMx® Spatial Informatics Platform (SIP) and in development and highlight results found in a publicly available colon adenocarcinoma FFPE WTX sample. First, we have redesigned the architecture with the goal of putting biology front and center, allowing researchers to quickly observe, hypothesize, and learn iteratively. Second, an agile computational engine, leveraging a foundational model trained on 1 billion spatially-resolved single cells, automates cell annotations, reducing analysis time from days to hours. Third, this workflow creates a carefully-curated package of summary tables, thousands of times smaller than a full dataset, but far richer than a human-generated prompt. Once an LLM ingests this package, it stands ready to answer diverse questions about the dataset, e.g. “what pathways are spatially variable in tumor cells?”, or “what are the spatial contexts T cells inhabit in this tissue?” We applied this approach to the colon sample (412,052 cells). Our foundational model automated cell typing and spatial domain assignment, providing the basis for the analysis. We then applied the LLM chat workflow. An initial query rapidly identified the dominant pro-tumoral signatures (EGFR, TGF-β, MAPK) and reduced pro-apoptotic TRAIL pathway. A follow-up “conversational” query for molecular drivers identified a significant MIF (tumor) to CD74 (TAM) ligand-receptor interaction, providing a direct mechanistic link to the observed MAPK activity. This entire workflow from automated cell typing to mechanistic insight was completed rapidly and generated all reproducible figures, demonstrating a complete analysis-to-visualization pipeline that does not require expert coding abilities. In conclusion, the AtoMx® SIP successfully bridges the gap between complex spatial data and biological insight. This new paradigm democratizes whole-transcriptome analysis, enabling researchers to conduct rapid, iterative, hypothesis-driven discovery without deep computational expertise. Citation Format: Evelyn R. Metzger, Patrick Danaher, Nicole Ortogero, Sayani Bhattacharjee, Prajan Divakar, Joseph M. Beechem. Scale meets Speed: A transcriptome-wide spatial analysis that runs a rapid and interactive hypothesis-driven workflow with AtoMx® SIP abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2757.