Abstract A033: Utilizing machine learning algorithms to identify methylation regions predictive of outcome in ICI-treated patients: Insights from the longitudinal RADIOHEAD study
Bibliographic record
Abstract
Abstract Background: Recent evidence suggests that tumor fraction (TF), estimated by normalizing cancer-specific differentially methylated regions against matched controls, is predictive of mortality and progression free survival (PFS) in cancer patients. However, specific methylation regions may be more predictive of outcomes than others. To investigate this assertion, we utilized a machine learning (ML) algorithm to identify regions that were most predictive of these outcomes while considering that the most predictive methylation regions may be different at various timepoints along the patient journey. Methods: RADIOHEAD is a prospective study of 1070 pan-cancer immunotherapy naive patients receiving standard of care immune checkpoint inhibitor (ICI) regimens in the community setting. Plasma samples from baseline and on-treatment timepoints from these patients were processed and a TF was reported by a clinically validated-methylation based next generation sequencing test (Guardant Reveal). Of these patients, a sub-cohort of 251 patients with advanced non-small cell lung cancer were analyzed further. Specifically, a gradient boosting (GB) model was used to identify the methylation regions most predictive of mortality and PFS at baseline and at three successive timepoints. Hyperparameters were optimized to maximize classification accuracy using a 3-fold cross-validation with 3 repeats. In total, eight GB models were leveraged, one for each outcome and timepoint. Results: The feature space, containing +10,000 regions, was interrogated using the GB models, where SHapley Additive exPlanations (SHAP) values are extracted. Through this examination, the top 10 methylation regions (those most accurately classified outcome) were identified for each model across all timepoints. Results are presented in tabular form where regions are ranked by importance. Conclusions: To our knowledge identifying the methylation regions that are most predictive of outcome using GB as a classifier is novel. In addition, we demonstrate that the most predictive regions differ over time. This process may be replicated in the future for different cancer types, where future research is needed to understand the biological significance of the identified regions. Citation Format: Christopher R. Pretz, Aaron Hardin, Bryan Lin, Sara Wienke, Samantha Liang, Enjun Yang, Amar K. Das. Utilizing machine learning algorithms to identify methylation regions predictive of outcome in ICI-treated patients: Insights from the longitudinal RADIOHEAD study [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 A033.
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 | 0.003 | 0.005 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| 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".