Many common human diseases are complex diseases. They are caused by not one single gene, but the interactions between large numbers of genes and environments. So the methods for Mendelian genetic disease study are not suitable for the complex disease very well. Undstanding the molecular mechanisms of complex diseases will be a great challenge in the medical and biological areas of the 21st century. With the end of Human Genome Project (HGP), we entered the postgenomic era. The studies based on HGP can be classified into three subjects: genome and biology, genome and health, genome and society. In this dissertation, we focus on the subject of genome and health.That is, from the point of bioinformatics for complex diseases, we aim to avail of complex disease related genomic, proteomic and genetic datasets, and develop effective bioinformtic methods to clarify the structure, function, and interactions of human genes and proteins, and the relationships between them and complex disease. The ultimate goal of our study is to develop the systematic methods for complex diseases, so as to understand the molecular mechanisms of complex diseases, and to provide valuable clues from their diagnosis and therapy. The main contents and contributions of this dissertation are as follows: 1. The classification of complex diseases based on microrray datasets: Most methods of pattern recognition and classification based on large-scale microarray datasets use one individual feature selection and classification method, we proposed a combinational feature selection and ensemble neural network method for classification of gene expression data. By combining various features, we can make full use of all available information, and can significantly improve the accuracy and stability of classification. On a wide range of published datasets, our method performs better, or is at least comparable to, the current best methods of our knowledge. 2. Complex disease gene finding based on systems biology: From the view of systems biology, we developed a new method to explore candidate genes for human brain diseases based on a brain-specific gene network. By integrating various diverse genomic and proteomic datasets based on Bayesian theory, we firstly constructed a complex human brain-specific gene network. Then we developed an effective method to find brain disease related gene subnetwork from the entire network. When this method is applied to predict Alzheimer’s disease related genes, the results show that this method can provide many valuable clues for other studies.
修改评论