We investigated the relationship of quantitative plaque features from coronary computed tomography (CT) angiography and coronary vascular dysfunction by impaired myocardial flow reserve (MFR) by (13)N-Ammonia positron emission tomography (PET). Fifty-one patients (32 men, 62.4±9.5 years) underwent combined rest-stress (13)N-ammonia PET and CT angiography scans by hybrid PET/CT. Regional MFR was measured from PET. From CT angiography, 153 arteries were evaluated by semiautomated software, computing arterial noncalcified plaque (NCP), low-density NCP (NCP<30 HU), calcified and total plaque volumes, and corresponding plaque burden (plaque volumex100%/vessel volume), stenosis, remodeling index, contrast density difference (maximum difference in luminal attenuation per unit area in the lesion), and plaque length. Quantitative stenosis, plaque burden, and myocardial mass were combined by boosted ensemble machine-learning algorithm into a composite risk score to predict impaired MFR (MFR≤2.0) by PET in each artery. Nineteen patients had impaired regional MFR in at least 1 territory (41/153 vessels). Patients with impaired regional MFR had higher arterial NCP (32.4% versus 17.2%), low-density NCP (7% versus 4%), and total plaque burden (37% versus 19.3%, P<0.02). In multivariable analysis with 10-fold cross-validation, NCP burden was the most significant predictor of impaired MFR (odds ratio, 1.35; P=0.021 for all). For prediction of impaired MFR with 10-fold cross-validation, receiver operating characteristics area under the curve for the composite score was 0.83 (95% confidence interval, 0.79-0.91) greater than for quantitative stenosis (0.66, 95% confidence interval, 0.57-0.76, P=0.005). Compared with stenosis, arterial NCP burden and a composite score combining quantitative stenosis and plaque burden from CT angiography significantly improves identification of downstream regional vascular dysfunction.