A machine learning model integrating circulating Temra cell transcriptional profiles to predict immunotherapy efficacy.
[BACKGROUND] While immunotherapy has transformed treatment paradigms for cancer, durable response rates remain below 30%.
APA
Dong Y, Xu X, et al. (2026). A machine learning model integrating circulating Temra cell transcriptional profiles to predict immunotherapy efficacy.. Med (New York, N.Y.), 7(4), 101027. https://doi.org/10.1016/j.medj.2026.101027
MLA
Dong Y, et al.. "A machine learning model integrating circulating Temra cell transcriptional profiles to predict immunotherapy efficacy.." Med (New York, N.Y.), vol. 7, no. 4, 2026, pp. 101027.
PMID
41747737
Abstract
[BACKGROUND] While immunotherapy has transformed treatment paradigms for cancer, durable response rates remain below 30%. Currently available biomarkers are insufficient for precise patient stratification. Blood-based biomarkers, as minimally invasive and accessible tools, have attracted attention for their potential to dynamically monitor immune checkpoint inhibitor response.
[METHODS] Paired peripheral blood and tumor samples from 15 patients were analyzed using single-cell RNA sequencing and T cell receptor sequencing. To construct a comprehensive CD8 T cell transcriptomic atlas, we integrated 13 publicly available single-cell datasets. A logistic regression-based model was trained on the transcriptional profiles of terminally differentiated effector memory CD8 T (Temra) cells and then validated in a prospective, multi-center non-small cell lung cancer (NSCLC) cohort undergoing first-line immunotherapy.
[FINDINGS] Circulating Temra cells were the most clonally expanded CD8 subset identified in peripheral blood. These cells exhibited distinct differentiation trajectories between responders and non-responders, with transcriptional profiles correlated with treatment efficacy. In the aggregated dataset comprising 748,082 CD8 T cells, Temra transcriptional signatures were consistently associated with therapeutic outcomes. The predictive model, built upon the transcriptional profiles of Temra cells, achieved high accuracy (87%-95%) across two external validation cohorts. Validation in a prospective cohort of 131 patients with advanced NSCLC (ClinicalTrials.gov: NCT06054152) confirmed strong predictive performance (area under the curve [AUC] = 0.834).
[CONCLUSIONS] Circulating Temra cells serve as robust predictors of immunotherapy efficacy. This study establishes a machine learning-based, pan-cancer, blood-derived transcriptomic biomarker that was prospectively validated in an NSCLC cohort and may substantially improve clinical decision-making in immuno-oncology.
[FUNDING] This study was funded by the National Natural Science Foundation of China (no. 92259205).
[METHODS] Paired peripheral blood and tumor samples from 15 patients were analyzed using single-cell RNA sequencing and T cell receptor sequencing. To construct a comprehensive CD8 T cell transcriptomic atlas, we integrated 13 publicly available single-cell datasets. A logistic regression-based model was trained on the transcriptional profiles of terminally differentiated effector memory CD8 T (Temra) cells and then validated in a prospective, multi-center non-small cell lung cancer (NSCLC) cohort undergoing first-line immunotherapy.
[FINDINGS] Circulating Temra cells were the most clonally expanded CD8 subset identified in peripheral blood. These cells exhibited distinct differentiation trajectories between responders and non-responders, with transcriptional profiles correlated with treatment efficacy. In the aggregated dataset comprising 748,082 CD8 T cells, Temra transcriptional signatures were consistently associated with therapeutic outcomes. The predictive model, built upon the transcriptional profiles of Temra cells, achieved high accuracy (87%-95%) across two external validation cohorts. Validation in a prospective cohort of 131 patients with advanced NSCLC (ClinicalTrials.gov: NCT06054152) confirmed strong predictive performance (area under the curve [AUC] = 0.834).
[CONCLUSIONS] Circulating Temra cells serve as robust predictors of immunotherapy efficacy. This study establishes a machine learning-based, pan-cancer, blood-derived transcriptomic biomarker that was prospectively validated in an NSCLC cohort and may substantially improve clinical decision-making in immuno-oncology.
[FUNDING] This study was funded by the National Natural Science Foundation of China (no. 92259205).
MeSH Terms
Humans; Machine Learning; Carcinoma, Non-Small-Cell Lung; Lung Neoplasms; CD8-Positive T-Lymphocytes; Prospective Studies; Immunotherapy; Male; Female; Middle Aged; Aged; Biomarkers, Tumor; Transcriptome; Immune Checkpoint Inhibitors; Treatment Outcome
같은 제1저자의 인용 많은 논문 (5)
- EIF4A3/circPTGR1/miR-4725-5p positive- feedback loop promotes colorectal cancer progression via FAK/AKT signaling pathway.
- Perioperative outcomes and postoperative functional outcomes of robot-assisted radical prostatectomy for oligometastatic versus localized prostate cancer: a multicenter matched case-control study.
- Germ layer specification and organotropism in lymphoma invasion.
- Pomegranate peel-derived exosome-like nanoparticles: A discarded treasure trove for colorectal cancer treatment.
- The fiber code: AI predicts HCC in MASLD.