Modeling Transcriptional Reprogramming by Markov Chain Monte Carlo Sampling

描述(申请人提供)：复杂的疾病，包括癌症和多发性硬化症，当多种生物系统受到分子变化的影响时，如基因突变、基因沉默或外部因素对细胞的修饰，就会发生。这些分子变化导致细胞状态的变化，几乎总是包括基因表达的重新编程。由于生物系统本质上是复杂的和非线性的，在许多情况下，需要对系统进行适当的建模来恢复系统级行为的知识。在这项提案中，我们的重点是提高我们从高通量数据推断生物过程活动的能力的方法，以及这些过程的生物标记物的开发。 我们开发了一种用于分析微阵列数据的马尔可夫链蒙特卡罗算法，该算法结合了转录调控知识和简单的信号网络模型，以识别癌症治疗药物的靶上和靶外效应。在这里，我们建议通过整合来自通路、基因突变、甲基化阵列、miRNA阵列和蛋白质-蛋白质相互作用(PPI)的数据来在系统水平上改进对细胞行为的推断，从而在该算法中包括先验知识。我们的第一个目标是开发一个与转录重编程相结合的集成信号模型，将路径知识与突变状态和miRNA水平的测量结合起来，以改进对细胞信号的估计。我们的第二个目标是将甲基化测量和转录因子结合位点信息作为先验数据来改进对转录因子靶标的推断，以改进对上游信号的基因集推断和估计。我们的第三个目标将开发一种方法学，根据基因的多重调控，从我们的模型中寻找可靠的生物标记物。 拟议工作的成功完成将为改进对细胞转录重编程的推断和生物标记物识别提供新的开源工具。这将大大改善对驱动系统脱离内稳态的具体变化的推断，以及对个别病例的治疗反应的推断。

DESCRIPTION (provided by applicant): Complex diseases, including cancer and multiple sclerosis, arise when multiple biological systems are impacted by molecular changes, such as gene mutation, gene silencing, or cellular modification by external agents. These molecular changes lead to changes in the cellular state, almost always including reprogramming of gene expression. Because biological systems are inherently complex and nonlinear, appropriate modeling of the system is required to recover knowledge of systems-level behavior in many cases. We focus in this proposal on methods to improve our ability to infer biological process activity from high- throughput data and on the development of biomarkers of these processes. We have developed a Markov chain Monte Carlo algorithm for analysis of microarray data that combines knowledge of transcriptional regulation and simple models of signaling networks to identify on-target and off- targets effects of therapeutics in cancer. Here we propose including prior knowledge in this algorithm by integrating data from pathways, gene mutations, methylation arrays, miRNA arrays, and protein-protein interactions (PPIs) to improve inference of cell behavior at a systems-level. Our first aim will develop an integrated model of signaling coupled to transcriptional reprogramming that couples knowledge of pathways to measurements of mutation status and miRNA levels to improve estimation of cell signaling. Our second aim will integrate methylation measurements and transcription factor binding site information as prior data to refine inference of targets of transcription factors to improve gene set inference and estimation of upstream signaling. Our third aim will develop a methodology to find reliable biomarkers from our models in light of multiple regulation of genes. Successful completion of the proposed work will provide novel open-source tools for improved inference on and biomarker identification of transcriptional reprogramming of cells. This will substantially improve inference of the specific changes that drive systems out of homeostasis and of treatment response in individual cases.