Physiopathological Mechanism of Alzheimer’s Disease
With the rapid development of neurobiology and neuroimaging technologies, mounting evidence shows that Alzheimer’s disease (AD) is caused by the build-up of two abnormal proteins, beta-amyloid and tau. In this project, we aim to improve our current understanding of the physiopathological mechanism in AD by offering a system-level mapping of how AD pathological burden spreads throughout the brain. We will develop a network-guided systems biology approach to characterize the dynamic reaction-diffusion process of AD pathology across brain networks and investigate the relationship between the insights of system behaviors and the trajectory of cognitive decline.
Deep Learning Techniques for Preterm Birth Prediction
Preterm birth (PTB) is a leading cause of long-term complications and mortality in children. Recent studies indicate that PTB is associated with transcriptome changes at different gestational ages. However, it is unknown whether there are distinct markers or signatures that can predict the risk of PTB. In this study, we will use the power of deep neural networks to understand preterm birth complications at the transcriptomic level and seek optimal predictive power. Specifically, we will first develop a deep learning model to predict the gestational age (GA) using maternal whole blood gene expression data. Second, we will generate diagnostic biomarker(s) to indicate PTB risk based on predictive model results.
Caulobactor Cell Cycle Modeling
Caulobacter is a bacterium that widely appears in freshwater. We are investigating stochastic spatiotemporal models of the Caulobacter cell cycle, which is a typical example of the asymmetric division in prokaryotic cells and has many remaining unknown mechanisms. In the project, both deterministic and stochastic systems were used to model the control mechanism, to analyze their interactions and to provide simulation results that can be compared with experimental observations in quantitative details.
Parameter Optimization of Biological Systems
Parameter estimation in stochastic cell cycle models is challenging since the number of simulations and the amount of empirical data must be large to obtain statistically valid parameter estimates. I worked with Dr. Watson, on parameter estimation of stochastic biological models using a new quasi-Newton stochastic optimization algorithm (QNSTOP). QNSTOP directly uses the random objective function value samples rather than creating ensemble statistics. Applying the algorithm to the budding yeast stochastic cell cycle model yielded predictions that matched well some summary statistics and one-dimensional distributions from empirical data.
Analysis of Hybrid Stochastic Algorithm
Stochastic effects in cellular systems are usually modeled and simulated with Gillespie’s stochastic simulation algorithm (SSA), but the low efficiency of SSA limits its application to large chemical networks. The hybrid method, proposed by Haseltine and Rawlings, can improve the simulation efficiency by combining ordinary differential equations for traditional deterministic models and SSA for stochastic models. we first showed that the hybrid method is accurate under certain thresholds and that it is valid for a much greater region in system parameter space compared with other approximate methods.