The human brain contains hundreds of regions with different anatomical structures and functions, and each region is composed of billions of heterogeneous cells. The expression pattern of the entire human brain transcriptome is extremely complex. At the microscopic level, the expression of thousands of individual genes determines the expression pattern of the genome through complex interactions. At the macro level, the interaction of genomic expression patterns among hundreds of different brain regions results in the transcriptional characteristics of the entire brain. At each level, there may involve paired, triploid, tetraploid, etc. interactions among genes or regions. Therefore, for an n-gene or n-region system, the complexity of its expression profile will be O(2n). However, the extremely high complexity limits traditional related methods. The transcription profile in a single brain area and that in an entire brain network is still unclear, which hinders the understanding of how gene expression regulates brain development, structure, function, and disease. Therefore, it is of great significance to find a simple quantitative framework to describe the structure of the whole genome expression pattern of a single brain area and that of an entire brain network. With this goal, by analyzing and modeling human brain transcriptome data, this article deeply studies the interaction structure that determines gene expression patterns, greatly simplifies the analysis framework of expression patterns, and on this basis, explores the application of this analysis framework in the research of brain disease-related genes.
Firstly, this paper finds that second-order interactions between genes determine the hierarchical transcription patterns in the human brain. The human brain is a complex system, whose structure and function are finely regulated by many genes. In addition, there are intricate interactions among genes. However, the basic rules of interactions among genes and their relationship with brain structure and function are unclear. By analyzing the transcription data of thousands of sampling sites and more than 10,000 genes in the human brain, it is found that the pairwise (second-order) interactions between gene expression can predict the transcription pattern of a single brain area and that of an entire brain network, indicating that the gene transcription pattern of the human brain is dominated by second-order interactions. This discovery greatly reduces the complexity of gene expression networks from O(2n) to O(n2). In addition, the study also revealed the potential relationship between gene expression and the brain network, that is, the strength of gene interaction observed empirically can lead to the nearly maximal number of transcriptional clusters, which may account for the functional and structural richness of the human brain during evolution.
Secondly, based on population coupling in the transcription, new organization and structures have been discovered at the level of genes and regions. Through further analysis of the transcriptome data of the whole brain, it is found that the correlation between the expression pattern of a single gene or region and that of the whole genome or brain network can capture pairwise interactions between genes or areas, which further reduces the complexity of the interaction structure of gene expression from O(n2) to O(n). According to population coupling, genes can be divided into strongly coupled "chorus" and weakly coupled "soloists". Both groups have different biological functions. Furthermore, at the region level, the study found obvious spatial clustering characteristics and region specificity according to the value of population coupling, suggesting an unreported organization of the human brain.
Finally, by using the population coupling of genes proposed above, this paper identifies genes whose interaction patterns change significantly in different brain diseases. Differentially expressed genes have been widely used to analyze the pathogenesis of human brain diseases. However, this analysis cannot identify genes with no significant changes in expression levels but altering their roles in the transcription network (for example, from "chorus" to "soloists"). Take brain diseases such as Alzheimer's (AD) as an example, through analyzing the RNA microarray dataset, it is found that in the comparison between the brain disease group and the normal control group, population coupling of many genes has changed. These genes are named role alternation genes (RAG) and only a minority of RAG are found to change significantly in gene expression level. Through gene enrichment analysis, this study clarified the biological significance of RAG. This work confirms that during the development of brain diseases, except for genes that change in expression levels, some genes will alter the way of interaction in the expression network, which will provide a new important perspective for the study of brain diseases.
In summary, this article uses parametric models to analyze the data of the human brain transcriptome, finds important factors that determine the structure of gene-gene interactions. Besides, this article proposes a series of dimensionality reduction methods under the premise of minimizing the loss of information, which greatly reduces the complexity when analyzing gene expression patterns. It also provides important theoretical foundation and analysis tools for subsequent related research. Further, by applying this innovative analysis framework to the study of brain disease-related genes, this article initially verifies its practical significance. We believe that the in-depth understanding of the structure of gene interaction will greatly promote the research on how genetic information determines the structure, function, development, and abnormality of brain networks in the future.
|Keyword||基因相互作用 功能连接 群体耦合 脑疾病|
|Sub direction classification||脑网络分析|
|华娇娇. 基于人脑转录组的基因相互作用结构研究[D]. 中国科学院自动化研究所. 中国科学院自动化研究所,2021.|
|Files in This Item:|
|Recommend this item|
|Export to Endnote|
|Similar articles in Google Scholar|
|Similar articles in Baidu academic|
|Similar articles in Bing Scholar|
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.