Madar Aviv, Bonneau Richard
Center for Comparative Functional Genomics, New York University, New York, NY, USA.
Methods Mol Biol. 2009;541:181. doi: 10.1007/978-1-59745-243-4_9.
Organisms must continually adapt to changing cellular and environmental factors (e.g., oxygen levels) by altering their gene expression patterns. At the same time, all organisms must have stable gene expression patterns that are robust to small fluctuations in environmental factors and genetic variation. Learning and characterizing the structure and dynamics of Regulatory Networks (RNs), on a whole-genome scale, is a key problem in systems biology. Here, we review the challenges associated with inferring RNs in a solely data-driven manner, concisely discuss the implications and contingencies of possible procedures that can be used, specifically focusing on one such procedure, the Inferelator. Importantly, the Inferelator explicitly models the temporal component of regulation, can learn the interactions between transcription factors and environmental factors, and attaches a statistically meaningful weight to every edge. The result of the Inferelator is a dynamical model of the RN that can be used to model the time-evolution of cell state.
生物体必须通过改变其基因表达模式来不断适应不断变化的细胞和环境因素(例如,氧气水平)。与此同时,所有生物体都必须具有稳定的基因表达模式,这些模式对于环境因素和遗传变异的小波动具有鲁棒性。在全基因组规模上学习和表征调控网络(RNs)的结构和动态,是系统生物学中的一个关键问题。在这里,我们回顾了以完全数据驱动的方式推断RNs所面临的挑战,简要讨论了可以使用的可能程序的含义和意外情况,特别关注其中一种程序,即Inferelator。重要的是,Inferelator明确地对调控的时间成分进行建模,可以学习转录因子和环境因素之间的相互作用,并为每条边赋予具有统计学意义的权重。Inferelator的结果是一个RN的动态模型,可用于模拟细胞状态的时间演变。