DSpace Repository

Statistical Frameworks for Integrative Analysis of Genetic Data

Show simple item record

dc.contributor.advisor Wang, Xuefeng en_US
dc.contributor.advisor Zhu, Wei en_US
dc.contributor.author Peng, Lizhen en_US
dc.contributor.other Department of Applied Mathematics and Statistics. en_US
dc.date.accessioned 2017-09-20T16:52:45Z
dc.date.available 2017-09-20T16:52:45Z
dc.date.issued 2015-05-01 en_US
dc.identifier.uri http://hdl.handle.net/11401/77466 en_US
dc.description 92 pg. en_US
dc.description.abstract We studied three major interconnected projects focused on establishing statistical frameworks for integrative analysis of genetic data, based on Cox proportional hazard models, kernel Cox regressions, multiple kernel learning models, regularized regression models, and support vector machines regressions, incorporated with dimensionality reduction and feature selections. (1) Our first project employed several machine learning algorithms for clinical predictions utilizing omics data across tumor types, to explore the potential benefits of including genetic measurements with traditional clinical information, for supporting doctor decisions. To predict survival of patients with tumors, our study focused on two objectives. First we applied multivariate Cox proportional hazard (Cox) models with univariate Cox screen or correlation screen, plus L1 penalized log partial likelihood (LASSO) for feature selection. Second, we also examined the factors that could affect prediction of dichotomized survival data by different machine learning algorithms, especially the MKL algorithms for it's capable of data fusion. Our analysis results indicate that incorporating Omics data with clinical information, can significantly improve predictions. Our also provided a well-established frameworks and resources, for reliable prognostic modeling and therapeutic decision making. (2) Our second project involved comprehensively assessing, by using genome-wide DNA methylation data as markers, the contribution of epigenetic effects on asthma and blood related quantitative traits. To evaluate the clinical utility of epigenetic markers, we constructed and compared various prediction models by including top ranked methylation loci from the genome-wide association scan, together with selected sets of known genetic markers from published genome-wide association studies. We observed a significant increase of correlation coefficient between actual and predicted IgE level when methylation markers were included. We also assessed the performance of cross platform prediction using methylation markers. Taken together, results from our assessment suggest that methylation has great potential in prediction of clinical phenotype. (3) Our third project explored kernel Cox regression models to improve the prediction accuracy of patients with metastatic castrate resistant prostate cancer (mCRPC) that treated by docetaxel. We proposed a future direction of utilizing clinical kernels in kernel Cox regression, to potentially obtain better results than linear and Gaussian kernels, with clinical variables only for prognostic modeling. en_US
dc.description.sponsorship This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree. en_US
dc.format Monograph en_US
dc.format.medium Electronic Resource en_US
dc.language.iso en_US en_US
dc.publisher The Graduate School, Stony Brook University: Stony Brook, NY. en_US
dc.subject.lcsh Statistics en_US
dc.subject.other Integrative Analysis, Statistical Frameworks en_US
dc.title Statistical Frameworks for Integrative Analysis of Genetic Data en_US
dc.type Dissertation en_US
dc.mimetype Application/PDF en_US
dc.contributor.committeemember Kuan, Pei Fen en_US
dc.contributor.committeemember Wu, Song en_US
dc.contributor.committeemember Bahou, Wadie. en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account