ROC Analysis for sEV Effectiveness Prediction Protocol
Phase 1: Sample Collection and sEV Isolation (Days 1-14)
Patient Recruitment: Enroll n=120 cardiovascular disease patients (60 with acute coronary syndrome, 60 age/sex-matched stable angina controls) undergoing elective cardiac catheterization at tertiary cardiovascular center. Obtain written informed consent for sEV collection from plasma.
sEV Isolation via Ultracentrifugation: Collect peripheral blood (20 mL EDTA) within 30 min of catheterization. Centrifuge at 300×g (10 min, 4°C), then 2000×g (10 min, 4°C) to remove cells and debris. Filter supernatant through 0.22 μm filter. Ultracentrifuge at 100,000×g (70 Ti rotor, 16 hours, 4°C). Resuspend pellet in 1 mL PBS, re-ultracentrifuge at 100,000×g (2 hours). Resuspend final pellet in 100 μL PBS. Store at -80°C.
Characterization: Verify sEV identity via NTA (NanoSight NS300), CD9/CD63/CD81 western blot, and transmission electron microscopy (100 nm scale). Exclude samples with >30% aggregated vesicles.
Phase 2: Biomarker Measurement (Days 15-28)
miR-130a Quantification: Extract sEV RNA via miRNeasy kit (Qiagen). Reverse transcribe miR-130a (hsa-miR-130a-3p, Qiagen RT primer) using TaqMan MicroRNA RT Kit. Run qPCR on QuantStudio 7 Flex (Applied Biosystems) with miR-130a-3p and snord48 endogenous control. Calculate ΔCt values.
TGF-β Content Analysis: Lyse sEVs in RIPA buffer, measure total TGF-β1 via ELISA (R&D Systems DY240). Normalize to sEV protein content (BCA assay). Run all samples in triplicate with internal reference standard.
Phase 3: ROC Curve Construction and Validation (Days 29-42)
Model Development: Split cohort 70/30 into derivation (n=84) and validation (n=36) sets. Build logistic regression model with miR-130a ΔCt, TGF-β1 concentration, and clinical covariates (age, sex, BMI, hypertension status). Evaluate single-marker and multi-marker models.
ROC Analysis: Generate ROC curves for each model. Calculate AUC with 95% DeLong CI. Determine optimal cutpoints via Youden index (J = sensitivity + specificity - 1). Assess calibration via Hosmer-Lemeshow test. Bootstrap validation (n=1000 resamples) for internal validation.
Clinical Utility Assessment: Calculate net reclassification improvement (NRI) and integrated discrimination improvement (IDI) comparing full model vs. clinical covariates alone. Decision curve analysis for threshold probabilities (0.1 to 0.9).