Discovering gene regulatory networks through combination of diverse large-scale biological evidence

当前后基因组分子生物学和生物信息学研究的一个主要挑战是充分理解基因调控：在细胞生命过程中，每个基因是如何以及何时开启和关闭的，特别是在响应不断变化的环境或细胞状态时。基因调控对生物体的生存至关重要，而破译基因是如何被调控的对于理解基因调控错误是至关重要的，而基因调控错误是疾病和发育畸形的主要原因。破译基因调控是一个逆向工程问题。目标是通过实验证据来理解一个系统（在这个例子中是细胞）是如何工作的。为此，生物学家干扰输入（环境变化或基因改变）并监测输出（例如，测量细胞中基因的相对丰度）。输出的不同测量代表了同一调节机制的各个方面，可以结合起来制定精确的基因调节模型。这些模型通常被称为基因调控或转录网络。计算方法可以利用正在生成的丰富的生物信息，并研究基因调控的一般原理。开发能够做到这一点的计算方法是一个非常活跃和有前途的研究领域。本提案的主要目标是开发一种新的机器学习系统，该系统结合了各种最近生成的生物学证据来推断转录网络。与该研究项目相关的预期结果是：1)提高我们对细胞如何控制基因表达的理解，这在人类疾病中起着关键作用；2)发现以前未知的调控相互作用；3)开发一种使能技术，使生物学家能够进一步分析自己的数据；4)开发新的计算机程序来模拟基因调控。该研究项目还将培养生物信息学跨学科领域的研究生。研究结果将引起生物信息学、计算机科学和生物学研究人员的兴趣，并且作为生物学家的使能技术，我们的系统将在加拿大和国外的分子生物学研究界产生重大影响。

A major challenge in current post-genome molecular biology and bioinformatics research is to fully understand gene regulation: how and when each and every gene is switched on and off during the life of the cell, particularly in response to changing environmental or cellular states. Gene regulation is essential to organism survival, and deciphering how genes are regulated is critical for understanding genetic misregulation, a major cause of disease and developmental malformations. Deciphering gene regulation is a reverse-engineering problem. The goal is to understand how a system (in this case the cell) works from experimental evidence. To this end, biologists disturb the inputs (changes in the environment or genetic alterations) and monitor the outputs (for example, measurements of relative abundance of genes in the cell). Distinct measurements of the outputs represent aspects of the same regulatory mechanism and can be combined to formulate accurate models of gene regulation. These models are usually called gene regulatory or transcriptional networks. Computational approaches can take advantage of the wealth of biological information that is being generated and investigate the general principles of gene regulation. Development of computational approaches capable of doing this is a very active and promising area of research. The main goal of this proposal is to develop a novel machine learning system that combines diverse recently-generated biological evidence to infer transcriptional networks. Expected outcomes associated with this research program are: 1) to improve our understanding of how cells control gene expression, which plays a key role in human disease; 2) to uncover previously unknown regulatory interactions; 3) to develop an enabling technology for biologists to further analyze their own data; and 4) to develop novel computer programs to model gene regulation. This research program will also train graduate students in the interdisciplinary area of bioinformatics. The research results will be of interest to researchers in bioinformatics, computer science and biology, and as an enabling technology for biologists our system will have a large impact in the molecular biology research community in Canada and abroad.