Abstract B040: Literature-scaled immunological gene set annotation using AI-powered immune cell knowledge graph (ICKG)
Bibliographic record
Abstract
Abstract Large scale application of single-cell and spatial omics in models and patient samples has led to the discovery of many novel gene sets, particularly those from an immunotherapeutic contexts. However, the biological meaning of those gene sets has been interpreted anecdotally through over- representation analysis against canonical annotation databases of limited complexity, granularity, and accuracy. Rich functional descriptions of individual genes in an immunological context exist in the literature but are not semantically summarized to perform gene set analysis. To overcome this limitation, we constructed immune cell knowledge graphs (ICKGs) by integrating over 24,000 published abstracts from recent literature using large language models (LLMs). ICKGs effectively integrate knowledge across individual, peer-reviewed studies, enabling accurate, verifiable graph- based reasoning. We validated the quality of ICKGs using functional omics data obtained independently from cytokine stimulation, CRISPR gene knock-out, and protein-protein interaction experiments. Powered through ICKG-based network analysis, we achieved rich, holistic, and accurate annotation of immunological gene sets, including those that were unannotated by existing approaches and those that are widely appreciated with clinical implications. We created an interactive website (https://kchen-lab.github.io/immune-knowledgegraph.github.io/) to perform ICKG-based immune gene set annotations and visualize the supporting rationale. Citation Format: Shan He, Vakul Mohanty, Yukun Tan, Qing Ye, Matthew Gubin, Hind Rafei, Weiyi Peng, Katy Rezvani, Ken Chen. Literature-scaled immunological gene set annotation using AI-powered immune cell knowledge graph (ICKG) [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 B040.
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 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.016 | 0.008 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.023 | 0.009 |
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".