Coronary heart disease (CHD) is one of the leading causes of mortality and morbidity in the United States. Accurate time-to-event CHD prediction models with high-dimensional DNA methylation and clinical features may assist with early prediction and intervention strategies. We developed a state-of-the-art deep learning autoencoder survival analysis model (AESurv) to effectively analyze high-dimensional blood DNA methylation features and traditional clinical risk factors by learning low-dimensional representation of participants for time-to-event CHD prediction. We demonstrated the utility of our model in two cohort studies: the Strong Heart Study cohort (SHS), a prospective cohort studying cardiovascular disease and its risk factors among American Indians adults; the Women's Health Initiative (WHI), a prospective cohort study including randomized clinical trials and observational study to improve postmenopausal women's health with one of the main focuses on cardiovascular disease. Our AESurv model effectively learned participant representations in low-dimensional latent space and achieved better model performance (concordance index-C index of 0.864 ± 0.009 and time-to-event mean area under the receiver operating characteristic curve-AUROC of 0.905 ± 0.009) than other survival analysis models (Cox proportional hazard, Cox proportional hazard deep neural network survival analysis, random survival forest, and gradient boosting survival analysis models) in the SHS. We further validated the AESurv model in WHI and also achieved the best model performance. The AESurv model can be used for accurate CHD prediction and assist health care professionals and patients to perform early intervention strategies. We suggest using AESurv model for future time-to-event CHD prediction based on DNA methylation features.

Bayesian nonparametric methods provide a convenient and well-founded framework for constructing spatio-temporally evolving ensemble models. They not only provide a flexible way to weight different underlying models in an ensemble according to the strengths of each model on different regions of the input space, but also allow for quantification of the ensemble’s uncertainty. However, computational costs can pose a challenge to kernel-based ensemble methods when spatio-temporal resolution is high, such as environmental models. We propose a Bayesian Nonparametric Ensemble (BNE) method for spatio-temporal ensemble learning based on the Gaussian process. Our ensemble relies on theoretically well-founded linearized approximations to the Gaussian process to adaptively weight the underlying models in a way that is scalable and amenable to stochastic learning methods. We investigate both the random Fourier feature and Nystrom approaches in this setting. We demonstrate the practicality and usefulness of the approximate model on the problem of air pollution prediction across the contiguous USA over a 6 year period.

Ensemble learning is a standard approach to building machine learning systems that capture complex phenomena in real-world data. An important aspect of these systems is the complete and valid quantification of model uncertainty. We introduce a Bayesian nonparametric ensemble (BNE) approach that augments an existing ensemble model to account for different sources of model uncertainty. BNE augments a model’s prediction and distribution functions using Bayesian nonparametric machinery. It has a theoretical guarantee in that it robustly estimates the uncertainty patterns in the data distribution, and can decompose its overall predictive uncertainty into distinct components that are due to different sources of noise and error. We show that our method achieves accurate uncertainty estimates under complex observational noise, and illustrate its real-world utility in terms of uncertainty decomposition and model bias detection for an ensemble in predict air pollution exposures in Eastern Massachusetts, USA.