Statistical Methods for ODE Models in AIDS Research

点击翻译按钮获取中文摘要

 DESCRIPTION (provided by applicant): Owing to the significant cost reduction of high-throughput technologies, frequent time course genome-wide gene expression data, in addition to time course cellular level and longitudinal phenotype response data, are often collected in recent HIV/AIDS studies and other biomedical projects. However, the effective use of the high-throughput time course data at transcriptomic and proteomics levels to study dynamic responses and network features is often hindered by lacking of statistical methods to reconstruct high-dimensional dynamic models. In this renewal project, we intend to fill this gap and propose the following specific aims: 1) Develop more efficient parameter estimation methods for high-dimensional ordinary differential equation (ODE) models. Aim 1 intends to develop more efficient statistical methods to estimate high-dimensional ODE model parameters to provide a foundation for reconstructing biological networks at gene, protein and molecular levels. 2) Develop novel statistical methods and implementation procedures for high-dimensional ODE variable selection to reconstruct the dynamic networks. We combine new statistical estimation methods for ODE models and regularization-based variable selection techniques to identify ODE network edges. Statistical methodologies and theoretical justifications will be established for the proposed ODE-based network models. 3) Evaluate and validate the methodologies developed in Aims 1-2 using computer simulations and real data analysis from HIV/AIDS studies. It is important to carefully evaluate the high-dimensional ODE variable selection and parameter estimation methods developed in Aims 1-2, and perform comparisons with existing methods for practical use. In particular, it is necessary to apply the proposed methods to experimental data from HIV/AIDS studies in order to demonstrate the usefulness of the proposed methodologies to address scientific questions. 4) Develop and disseminate efficient computational algorithms and user-friendly software tools for the proposed methods to the broader research community. It is very important to develop efficient computing algorithms and share/disseminate the computational source codes to the general research community.