Personalized risk predictor for acute cellular rejection in lung transplant using soluble CD31

Area of Science:

• Transplantation immunology
• Artificial intelligence in medicine
• Biomarker discovery

Background:

• Acute cellular rejection (ACR) is a major complication following lung transplantation (LTx).
• Early detection of ACR is crucial for patient outcomes.
• Soluble CD31 (sCD31) is a potential biomarker involved in immune regulation at the blood-vessel interface.

Purpose of the Study:

• To evaluate the utility of artificial intelligence (AI) in predicting ACR risk post-LTx.
• To investigate the combined predictive value of early plasma sCD31 levels, arterial oxygen partial pressure to fractional inspired oxygen ratio (PaO2/FiO2), and respiratory Sequential Organ Failure Assessment (SOFA) score.

Main Methods:

• A cohort of 40 lung transplant recipients was studied.
• Plasma sCD31 levels, PaO2/FiO2, and respiratory SOFA scores were measured within 3 days of LTx.
• Multivariate and multimodal models were used to analyze the nonlinear relationship between sCD31, PaO2/FiO2, and respiratory SOFA.
• A deep convolutional neural network was employed to classify time-series data and predict ACR risk.

Main Results:

• Seven out of 40 recipients (17.5%) experienced ACR.
• The study successfully modeled the dynamic interplay of sCD31, PaO2/FiO2, and respiratory SOFA.
• AI-driven classification identified individuals at risk for ACR based on integrated biomarker and physiological data.

Conclusions:

• AI, utilizing sCD31 and respiratory parameters, shows promise in predicting ACR risk after lung transplantation.
• This approach may facilitate earlier intervention and improved management strategies for LTx recipients.
• Further validation in larger cohorts is warranted to confirm the clinical utility of this AI-based predictive model.