Modeling Core

项目总结（见说明）： 建模核心将整合项目1和2的数据，并通过基于回归的推断生物体间对基因表达的直接和间接影响，构建联合宿主和病原体基因调控网络（GRN）模型。这种跨物种GRN模型将把定量表型与影响结核感染结果的因果环境和遗传触发因素联系起来，特别是易感性、耐药性、感染进展、持久性和清除。 这些预测模型的核心是机器学习算法-- cMonkey和Inferelator --它们将协同工作，以发现受组合环境（例如pH）和遗传（例如转录因子）影响的条件性共调控的基因组。纳入这些算法的数据整合策略将克服几个挑战：（1）系统生物学数据中的技术和生物噪声;（2）基因组中超过50%的基因缺乏功能信息;（3）缺乏对调控机制的详细了解;（4）对宿主和MTB网络都适应的环境空间的不完整了解。建模核心将直接从数据中逆向工程GRNs的架构，同时学习转录调控的相关动态。此外，未知功能的基因将基于其共表达模式和其他共享特征（如相互作用、同源性和顺式调控元件）整合到网络中;从而有可能发现可能对感染结果至关重要的其他基因。 已经构建了MTB和BMMO的初步GRN模型，对这些模型的分析表明，它们概括了现有的知识，并预测了可能影响MTB感染结果的新基因。这些模型预测为两个项目的实验设计和优先级提供了指导。

PROJECT SUMMARY (See instructions): The Modeling Core will integrate data from Projects 1 and 2 and construct a joint host and pathogen gene regulatory network (GRN) model by regression-based inference of direct and indirect inter-organismal influences on gene expression. This cross-species GRN model will link quantitative phenotypes to causal environmental and genetic triggers that influence the outcome of TB infection, specifically susceptibility, resistance, infection progression, persistence, and clearance. Central to these predictive models are machine learning algorithms - cMonkey and Inferelator - that will work in tandem to discover groups of genes that are conditionally co-regulated by combinatorial environmental (e.g. pH) and genetic (e.g; transcription factors) influences. The data integration strategies incorporated into these algorithms will overcome several challenges: (1) technical and biological noise in systems biology data; (2) lack of functional information for over 50% of all genes in the genome; (3) lack of detailed knowledge of regulatory mechanisms; and (4) incomplete knowledge of the environmental space to which both the host and MTB networks have adapted. The Modeling Core will reverse engineer the architecture of GRNs directly from data while simultaneously learning the associated dynamics of transcriptional regulation. Moreover, genes of unknown function will be integrated into the network based on their co-expression patterns, and other shared features such as interactions, phylogeny, and cis-regulatory elements; making it possible to discover additional genes that might be critical for outcome of infection. Preliminary GRN models for both MTB and BMMO have already been constructed, and analysis of these models has demonstrated that they recapitulate existing knowledge and predict new genes that might, influence the outcome of MTB infection. These model predictions have provided guidance for experimental designs and priorities in both Projects.