Abstract 5552: Uncovering clinically significant tumor microenvironment interaction programs across diverse cancers
Bibliographic record
Abstract
Abstract Background: Cell-cell interactions (CCI) within the tumor microenvironment (TME) play pivotal roles in various tumor behaviors, including cancer growth, metastasis, immune evasion, and resistance to therapy. Compared to front-line treatments, targeting specific CCIs marks a paradigm shift in cancer therapy, aiming for enhanced response rates with reduced side effects. Although current approaches targeting immune regulation pathways exhibit effective anti-tumor responses, substantial variability in response rates across patients and cancer types persists. Consequently, a critical imperative emerges to discover novel interaction programs (IPs) across diverse TMEs, demonstrating clear clinical impact across a broad spectrum of patients. Methods: Characterization of pan cancer IPs. We compiled a pan cancer single-cell RNA-seq (scRNA-seq) atlas, including 4 million cells from 890 tumors across 14 cancer types. In each sample, CCIs were inferred using a ligand-receptor (LR) analysis framework, generating ranked CCI scores based on consensus from various LR inference methods. To find IPs enriched across TMEs, we applied an unsupervised factorization approach. The outcome represents CCIs as factors, where sample loading indicates factors’ strength, and feature loading highlights the cells and LR pairs constituting each IP. Evaluating IPs in clinical RNA-seq cohorts. Given the more immediate translational potential of bulk cancer sequencing for patient stratification, we devised an approach to predict IPs' strengths in bulk RNA-seq. Random Forest regression models using pseudo-bulk expression of cancer genes were trained to predict the loading (strength) of factors (IPs) in each sample in the scRNA-seq pan-cancer atlas. These models were subsequently applied to cancer samples in The Cancer Genome Atlas (TCGA) cohort to assess the clinical impact of various IPs on patient mortality. Results: We assessed the impact of IPs on 5-year overall survival using Cox proportional hazard models. We identified multiple factors significantly associated with survival in at least 4 cancer types. Notably, a robust prognostic factor linked to the inclusion of anti-tumor natural killer cell interactions and exclusion of pro-tumor macrophage interactions was found in MESO, LGG, ACC, and UVM (HR 1.6-6.6, p < 0.05). The validation of IPs using spatial transcriptomics and evaluation of IP prediction's utility in targeted treatment cohorts are currently underway. Conclusion: We highlight a potent method for detecting clinically significant IPs involving cells in TME communicating through LR pairs. The ability to assess these scRNA-seq-derived IPs in clinical bulk RNA-seq cohorts signifies a valuable advancement. Our approach holds promise for expanding the repertoire of cell therapies, identifying new treatment targets, and improving our predictive capabilities for responses to existing therapies. Citation Format: Ido Nofech-Mozes, Vivian Wang, Philip Awadalla, Sagi Abelson. Uncovering clinically significant tumor microenvironment interaction programs across diverse cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5552.
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 | 0.001 |
| 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 | 0.000 | 0.000 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 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".