Font Size: a A A

Developing Statistical Approaches To Detecting Gene-Gene Interaction

Posted on:2010-09-26Degree:DoctorType:Dissertation
Country:ChinaCandidate:G B ChenFull Text:PDF
GTID:1100360302958848Subject:Crop Genetics and Breeding
Abstract/Summary:PDF Full Text Request
This dissertation concentrates on developing nonparametric method to detect gene-gene interactions underlying complex traits.In order to facility real data analysis,bioinformatics software GMDR(Generalized Multifactor Dimensionality Reduction),which realized several different versions of multifactor dimensionality reduction methods,is developed to help geneticists to hunt interacting genes.As the significant of tobacco in both agronomy and health care,in real cases,tobacco relevant data,such as a nicotine dependency cohort,will be utilized to demonstrate the capability of the methods proposed.This dissertation consists of five chapters.Chapter 1 is an overall introduction.Chapter 2 identified that APBB1 was associated to Nicotine Dependency(ND) under three nicotine measures assessed by SQ,HSI and FTND.Five Single Nucleotide Polymorphisms(SNPs) were selected and genotyped,rs4758416 was significantly associated to ND after the correction of multiple testing.In haplotype analysis,C-G-T and G-G-T, consisted by rs4758416-rs10839562-rs1079199,were significantly associated with ND. Another haplotype,formed by rs10839562-rs1079199-rs8164,was significantly associated with HSI too.Chapter 3 introduced family-based generalized multifactor dimensionality reduction (FGMDR).The new built algorithm is able to handle both discrete and quantitative traits and allows adjustment of covariates with family-based design.Compared with pedigree-based GMDR(PGMDR),FGMDR alleviates the computation burden but yields similar statistical power.Chapter 4 tried to establish the guideline for investigating the design issue of how to detect interaction in genetic epidemiological studies The results show that GMDR with covariate adjustment could give a power of>50%under a case-control design with a sample size of 1000~2,000,when the theoretical accuracy ranges from 0.56 to 0.65. Chapter 5 illustrated GMDR software.Currently,there are three methods included:1) GMDR,2) PGMDR,3) FGMDR.In order to facility the application,a Perl script is written to streamline the whole procedure.
Keywords/Search Tags:Statistical
PDF Full Text Request
Related items