In vitro machine learning-based CAR T immunological synapse quality measurements correlate with patient clinical...

Area of Science:

• Immunology and Artificial Intelligence
• Cancer Immunotherapy
• Machine Learning in Medicine

Background:

• Chimeric antigen receptor (CAR)-T cell immunotherapies have transformed blood cancer treatment.
• Assessing CAR-T cell efficacy and predicting patient response is challenging due to complex immunological synapse (IS) formation.
• Current methods for evaluating IS quality are time-consuming, labor-intensive, and lack accuracy and reproducibility.

Purpose of the Study:

• To develop a novel, rapid, and accurate machine learning (ML)-based method for quantifying CAR-T cell immunological synapse (IS) quality.
• To correlate ML-quantified CAR IS quality with clinical outcomes in patients treated with CAR-T cell therapy.
• To establish CAR IS quality as a potential predictive biomarker for CAR immunotherapy response.

Main Methods:

• Utilized a glass-support planar lipid bilayer system to form CAR-T cell IS.
• Developed and applied a novel artificial neural network (ANN) model incorporating object detection for image segmentation and quantification of thousands of CAR IS images.
• Correlated the ML-quantified CAR IS data with clinical response data from patients treated with Kappa-CAR-T cells.

Main Results:

• The ML-based automated algorithm accurately and rapidly quantified CAR-T IS quality.
• ML-quantified CAR IS quality demonstrated a strong correlation with clinical response (responder vs. non-responder) in patients.
• The developed method provides a quantitative measure of CAR IS quality that is discriminative for predicting patient response.

Conclusions:

• CAR cell IS quality, quantified using ML, serves as a potential composite biomarker for predicting antitumor activity and response to CAR immunotherapy.
• This ML-based approach offers a more efficient and accurate alternative to conventional methods for evaluating CAR-T cell efficacy.
• The developed methodology can guide the design and optimization of novel CAR constructs for improved clinical applications in cancer treatment.