A machine learning model integrating circulating Temra cell transcriptional profiles to predict immunotherapy...

Area of Science:

• Immunology
• Oncology
• Genomics

Background:

• Immunotherapy has revolutionized cancer treatment but durable response rates remain low (<30%).
• Current biomarkers lack precision for patient stratification.
• Blood-based biomarkers offer a minimally invasive approach to monitor treatment response dynamically.

Purpose of the Study:

• To identify and validate blood-based biomarkers for predicting immunotherapy efficacy.
• To develop a machine learning model using transcriptomic data of specific immune cells.
• To assess the predictive performance of these biomarkers in a prospective cohort.

Main Methods:

• Single-cell RNA and T cell receptor sequencing of paired blood and tumor samples.
• Integration of 13 public single-cell datasets to create a CD8+ T cell atlas.
• Training and validation of a logistic regression model on Temra cell transcriptional profiles.

Main Results:

• Circulating Temra cells showed the highest clonal expansion in peripheral blood.
• Temra cell transcriptional signatures correlated with treatment efficacy and patient outcomes.
• The predictive model achieved 87%-95% accuracy in external validation and an AUC of 0.834 in a prospective NSCLC cohort.

Conclusions:

• Circulating Temra cells are reliable predictors of immunotherapy efficacy.
• A machine learning-based, pan-cancer, blood-derived transcriptomic biomarker was established.
• This biomarker shows potential to significantly improve clinical decision-making in immuno-oncology.