Abstract B044: Distributed Transcriptomic Modeling for Biomarker Discovery in Immuno-Oncology
Bibliographic record
Abstract
Abstract Background: Immuno-oncology (IO) therapies, particularly immune checkpoint inhibitors, have transformed cancer care but benefit only a subset of patients. Existing biomarkers such as PD-L1 expression, tumor mutational burden, and microsatellite instability have limited predictive power and are often cancer-type specific. Efforts to identify more generalizable biomarkers are hampered by tumor heterogeneity, restricted access to high-quality datasets, and stringent data privacy regulations, making traditional centralized discovery approaches difficult to scale. To address these challenges, we developed a distributed pipeline for discovering predictive transcriptomic biomarkers across diverse clinical datasets without sharing sensitive patient data. Methods: We developed a scalable, cloud-based distributed pipeline for biomarker discovery that operates across institutions without transferring sensitive patient data. Using curated gene expression profiles, we assessed tumor microenvironment and IO-related signatures in pan-cancer, cancer-specific, and treatment-specific contexts. A multivariable XGBoost model was trained within this distributed framework while maintaining local data governance. Model performance was evaluated against univariable signature-based models and multivariable models trained locally on individual datasets. Results: We implemented a distributed meta-integration pipeline to 18 datasets comprising 967 patients across seven cancer types treated with PD-1/PD-L1, CTLA-4, or combination immunotherapies. This analysis identified biomarkers specific to lung cancer and CTLA-4 therapy that were not detected in pan-cancer analyses, highlighting the value of context-specific associations. Several gene expression signatures were consistently linked to progression-free survival and treatment response in both pan-cancer and melanoma datasets, with notable overlap between PD-1/PD-L1 and dual checkpoint blockade therapies. Distributed multivariable models outperformed locally trained models on average, indicating improved generalizability, though performance often depended on cancer type and dataset composition, with univariable signature-based models outperforming in some settings. Conclusion: This study demonstrates a distributed framework for IO biomarker discovery that preserves patient privacy and facilitates collaboration across institutions. Our findings highlight the value of multi-institutional, diverse datasets in refining context-specific biomarkers and advancing precision IO therapies. Citation Format: Farnoosh Abbas-Aghababazadeh, Kewei Ni, Kevin (Xin) Wang, Sisira Kadambat Nair, Nasim BondarSahebi, John Stagg, Benjamin Haibe-Kains. Distributed Transcriptomic Modeling for Biomarker Discovery in Immuno-Oncology [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 B044.
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 | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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".