Genomic Network Tomography

在活细胞中，由蛋白质-蛋白质相互作用组成的信号通路将细胞外条件的信息从细胞壁传递到细胞核，导致基因及其蛋白质产物表达的变化，使细胞能够适应不同的环境并生存。生物学家已经发现了某些信号通路的部分内容，但目前对完整信号网络结构的理解还远未完成。由于与体内测量相关的固有困难，本研究考虑使用现有高通量实验生成的数据推断细胞信号网络结构的问题，这些数据表明每个信号通路中使用了哪些蛋白质。细胞信号网络是活细胞生长、发育和生存的基础，因此该项目的结果可能会提高人类疾病、生物传感器开发和生物燃料制造等关键领域的知识水平。该项目研究了一种重建细胞信号网络的新技术，该技术基于现有高通量实验生成的数据，这些数据表明每个信号通路中使用了哪些蛋白质，但不直接揭示通路的结构/顺序。细胞信号网络和实验数据通过打乱的马尔可夫过程进行数学建模，这解释了数据不揭示通路结构/顺序的事实。混洗马尔可夫模型将网络重构问题简化为推断马尔可夫转移矩阵的任务。为此任务开发了基于期望最大化和重要性采样技术的计算高效的推理算法。使用真实和合成生物数据以及数学分析技术的计算实验证明了模型和算法的功能。

In living cells, signaling pathways composed of protein-protein interactions communicate information about extracellular conditions from the cell wall to the nucleus, leading to changes in the expression of genes and their protein products that enable the cell to adapt and survive in diverse environments. Biologists have uncovered portions of certain signaling pathways, but the current understanding of full signaling network structures is far from complete. Due to intrinsic difficulties associated with in vivo measurement, this research considers the problem of inferring the structure of a cellular signaling network using data generated by existing high-throughput experiments that indicate which proteins are utilized in each signaling pathway. Cell signaling networks underlie the growth, development and survival of living cells, and therefore the results of this project may advance the state of knowledge in the critical areas of human disease, biosensor development, and biofuel manufacturing.This project investigates a new technique for the reconstruction of cell signaling networks that is based on data generated by existing high-throughput experiments that indicate which proteins are utilized in each signaling pathway, but do not directly reveal the structure/order of the pathways. The cell signaling networks and the experimental data are mathematically modeled by a shuffled Markov process, which accounts for the fact that the data do not reveal the pathway structure/order. The shuffled Markov model reduces the network reconstruction problem to the task of inferring the Markov transition matrix. Computationally efficient inference algorithms, based on expectation-maximization and importance sampling techniques, are developed for this task. Computational experiments using real and synthetic biological data, as well as mathematical analysis techniques, demonstrate the capabilities of the model and algorithms.