Identifying Both Common And Individual Copy Number Variations Through Multiple Constrained Optimization

Copy number variant(CNV)is a novel structural variation ranging from 1KB to 3MB in human chromosomes,which mainly caused by gene rearrangement.CNV is an important component of gene structure variation that plays an important role in researching various complex diseases.Many researchers proposed different approaches for the identifying of copy number variants,such as CNV-TV and Penncnv.However,these approaches can only analysis the CNVs of a single sample at a time,and the major etiological factor may occupy only a fraction of all the detected CNVs,so it is difficult to find the CNVs linked with particular diseases using these methods.Meanwhile,patients with the same diseases may have different CNVs due to the specificity,so it is also a key problem to distinguish the individual CNVs of each sample.Based on this fact,we propose two different approaches to distinguish the individual CNVs and group CNVs;The main works of this thesis are as follows):Firstly,this thesis introduces some basic theories for analyzing CNVs and some approaches to distinguish between the individual CNVs and group CNVs And then,based on the sparsity,smoothness and structural similarity of group CNVs as well as the sparsity and smoothness of individual CNVs,we propose a multi-norm constrained model,named JNCO,to distinguish between the individual CNVs and group CNVs and also efficient coordinate-cyclic algorithms have been developed for based on this strategy.What's more,because of the strong ability of wavelet in denoising and local feature description,we propose a framework for detecting individual CNVs and group CNVs based on wavelet transform.On this foundation,we analysis the characteristics of individual CNVs and group CNVs after wavelet transform and then set up a wavelet transform based individual CNVs and group CNVs detecting model WaveDec.Finally,simulated data sets and real data sets are employed to evaluate the performance of these methods.The results show that our approaches are very competitive with other state-of-art approaches.In conclusion,the main job and achievement are as follow: 1)This paper analyzes the dynamic characteristics between different norms and individual CNVs or group CNVs,and then develop the JNCO model to distinguish between individual CNVs and group CNVs by using the appropriate norm.2)Depending on the strong abilities of wavelet in denoising and local feature description,we propose a distinguishable frame based on wavelet transform,and we also refines the WaveDec according to the characteristics of group and individual.3)The reliability of the proposed methods are proved by simulation and real experiments.