Type
Text
Type
Dissertation
Advisor
Zhu, Wei | Yang, Jie | Wu, Song | Ju, Jingfang.
Date
2015-05-01
Keywords
Statistics | miRNA, nonlinear mixed model, RPPA
Department
Department of Applied Mathematics and Statistics.
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/77499
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
Understanding functions of microRNAs (or miRNAs), particularly their effects on protein degradation, is biologically important. Emerging technologies, including the reverse-phase protein array (RPPA) for quantifying protein concentration and RNA-seq for quantifying miRNA expression, provide a unique opportunity to study miRNA-protein regulatory mechanisms. A naïve and commonly used way to analyze such data is to directly examine the correlation between the raw miRNA measurements and protein concentrations estimated from RPPA through simple linear regression models. However, the uncertainty associated with protein concentration estimates is ignored, which may lead to less accurate results and significant power loss. Here we propose an integrated nonlinear hierarchical model for detecting miRNA targets through original RPPA intensity data. The model is fitted within a maximum likelihood framework and the significance of the correlation between miRNA and protein is assessed using the Wald test. Our extensive simulation studies demonstrated that the integrated method performed consistently better than the simple method, especially when the RPPA intensity levels are close to the boundaries of image intensity limits. The proposed model was also illustrated through real datasets from The Cancer Genome Atlas (TCGA) program. In addition, we extend the model to a semi-parameter model by incorporating a nonparametric curve fitting technique, which relaxes the assumption of a specific parametric form for the RPPA response curve. The performance of this model is also demonstrated by simulation studies and real data analyses. | Understanding functions of microRNAs (or miRNAs), particularly their effects on protein degradation, is biologically important. Emerging technologies, including the reverse-phase protein array (RPPA) for quantifying protein concentration and RNA-seq for quantifying miRNA expression, provide a unique opportunity to study miRNA-protein regulatory mechanisms. A naïve and commonly used way to analyze such data is to directly examine the correlation between the raw miRNA measurements and protein concentrations estimated from RPPA through simple linear regression models. However, the uncertainty associated with protein concentration estimates is ignored, which may lead to less accurate results and significant power loss. Here we propose an integrated nonlinear hierarchical model for detecting miRNA targets through original RPPA intensity data. The model is fitted within a maximum likelihood framework and the significance of the correlation between miRNA and protein is assessed using the Wald test. Our extensive simulation studies demonstrated that the integrated method performed consistently better than the simple method, especially when the RPPA intensity levels are close to the boundaries of image intensity limits. The proposed model was also illustrated through real datasets from The Cancer Genome Atlas (TCGA) program. In addition, we extend the model to a semi-parameter model by incorporating a nonparametric curve fitting technique, which relaxes the assumption of a specific parametric form for the RPPA response curve. The performance of this model is also demonstrated by simulation studies and real data analyses. | 110 pages
Recommended Citation
Zhu, Jiawen, "MicroRNA Target Identification by Reverse Phase Protein Array" (2015). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 3311.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/3311