Genetic Association Analysis Of Rare Variants

Genome-wide association study (GWAS) is widely applied in the multiple com-plex disease/traits genetic studies, and becomes the main strategy and tool for associa-tion studies between gene and complex diseases for the moment, along with the com-pletion work of gene sequencing. Although immense success have been achieved in dissecting the genetic basis of complex diseases by GWAS, for most common diseases the proportion of the overall phenotype variance explained by the discovered disease-susceptibility found by the GWAS remains only5%-10%at most, and the heredity explanatory ability (heritability) of disease gene locus by the GWAS nearly reaches it’s limit. Based on this, people start to pay attention to solve the limitations of GWAS applying to association studies between the gene and disease traits. In this thesis, we study the association test problem between rare variants and disease traits under the common disease rare variant (CDRV) hypothesis.The premise hypothesis of the GWAS is the common disease common variants (CDCV) hypothesis, and the main study target of the GWAS is common gene vari-ants with the minor allele frequency (MAF) more than0.05commonly. The low level of the association analysis heritability under the CDCV hypothesis promotes people start to think about the common disease rare variants (CDRV) hypothesis, which stud-ies the association between rare mutations and disease. The rare variants refer to the rare causal variants with minor allele frequency less than or equal to1%but each with modest effect residing within a same functional unit, such as gene、pathway or ultra conserved regions. Generally speaking, the minor allele frequency of the rare variants although probably with high effect size, is small, leading to a less information com-posed by rare variants. Besides, the mechanism of rare variants to the disease traits is that there are many different variants across the entire gene having an independent ef-fect on the disease risk but each variant contribute just a small fraction, so the power of detection the association between gene and the disease traits always squint towards on the low side. The traditional association analysis tests include Single-Marker Tests and Multiple-Marker Tests, both have low power in rare variants and gene traits association studies, so are not applicable to the rare mutation test problem.In this paper, we propose two different test methods for the association test prob-lem between rare variants and the disease traits:Risk-based Variable-threshold Method (rb-VT) and Poisson Approximation based Score Test (PAST). The two methods both apply the idea of pooling the information together, pool the rare variants together to solve the rareness problem of rare variants, test the association problem as well as choose the variants. The Risk-based Variable-threshold Method applies the multi-factorial dimensionality reduction method theory, uses the risk factor of the case sam-ples to the control samples as criterion, combine the risk genes and gets the maximal value of a serious of test statistics by changing the threshold choices, and gets the sig-nificance level by permutation; the Poisson Approximation based Score Test treats the association test problem between rare variants and the disease traits from a different point of view of rare events, reanalyzes with poisson distribution approximating the frequency of multiple rare mutations, derives a score test statistic approximately as the standard normal distribution under the null hypothesis, and also reorders the gene variants using the risk factor of the case samples to the control samples as criterion to get the maximal value of a serious of test statistics, and gets the significance level by permutation ultimately. We use massive simulation tests to compare with variable-threshold approach and weighted-sum method, the results show that our methods both raise the power of the association test, and can choose the disease variants more ef-fectively. The two new test approaches provide new way for the association analysis method between rare variants and the disease traits.