Abstract B050: Spatially aware transcriptomic topic modeling reveals novel signals of spatial organization in glioblastoma
Bibliographic record
Abstract
Abstract Background: Spatial transcriptomics (ST) enables high-resolution mapping of gene expression across tissue architecture and can reveal the organization of cellular states in complex tissues, such as the tumor microenvironment. Using ST data from glioblastoma (GBM) patients, a large recent study1 identified robust gene programs in a spatially oblivious manner and used them to assign cell states to spatial spots and assess cell-state spatial organization. The results characterized GBM tumors as generally spatially heterogeneous, with some hypoxia-associated structured regions. However, recent computational methods support analyses that more tightly integrate spatial context and gene expression to identify continuous gradients and rare niches in ST data. We hypothesized that applying such an approach may reveal additional key aspects of the spatial organization of transcriptional programs in GBM tumors. Methods: Based on our previous successes with traditional topic modeling in transcriptomic analysis, we focused on the recently published method Spatial Transcriptomics Analysis with topic Modeling to uncover spatial Patterns (STAMP)2. STAMP uses a deep generative model to perform spatially aware topic modeling that can account for replicates. Using STAMP, we reanalyzed the 26 ST samples (10x Visium, 60,000+ spots) from 17 patients in the GBM dataset1 to infer spatial topics or gene programs, which we then interpreted in the context of previous studies. We also extended the analysis framework to enable quantification and statistical assessment of the spatial coherence of continuous gene program weights. Results: Consistent with previous studies, we identified gene programs associated with a hypoxic niche, as well as with precursor and differentiated endothelial, oligodendrocyte and neuronal cell types. However, our study also revealed a number of new gene programs with strongly spatially organized patterns of expression occurring in multiple patient samples. These findings indicate additional spatial axes that may capture important variation in and across GBM tumor microenvironments. Conclusion: Complementing a previous hypoxia-associated model of GBM spatial organization, our study characterized additional, spatially resolved functional processes, which may be clinically and therapeutically relevant. The results suggest that a spatially aware, non-categorical approach to ST analysis may drive discovery of previously hidden, spatially coherent, recurrent gene programs in the tumor microenvironment. This framework is broadly applicable for dissecting the transcriptionally governed spatial architecture of solid tumors. References: 1Greenwald, A.C. et al. Integrative spatial analysis reveals a multi-layered organization of glioblastoma. Cell 187, 2485–2501.e26 (2024). 2ZHong, C., Ang, K.S. & Chen, J. Interpretable spatially aware dimension reduction of spatial transcriptomics with STAMP. Nat. Methods 21, 2072–2083 (2024). Citation Format: Joseph J. Sifakis, Sophia Madejski, Samantha J. Riesenfeld. Spatially aware transcriptomic topic modeling reveals novel signals of spatial organization in glioblastoma [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Artificial Intelligence and Machine Learning; 2025 Jul 10-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(13_Suppl):Abstract nr B050.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".