Tensor Computations for Modeling Large-Scale Molecular Biological Data

描述（由申请人提供）：我们的研究受到高通量技术的最新进展的推动，例如DNA微阵列，这使得记录指导细胞过程进展的完整基因组信号成为可能。未来在生物学和医学领域的预测能力和发现将来自对这些快速增长的大规模分子生物学数据集的数学建模。为此，我们利用矩阵计算的框架建立了第一个数据驱动的基因组数据预测模型。我们通过分析细胞周期表达数据、转录因子和复制起始蛋白的dna结合数据等来说明这些模型。我们的模型预测以前未知的生物学原理的能力，通过预测酵母中复制起始与细胞周期调节转录相关的一种新的调节机制得到了证明。现在，我们的目标是通过收集和分析真核生物中被认为与细胞周期转录解耦的条件下的全基因组表达，在实验上验证这一计算预测。这些实验将检验我们的数学模型正确预测生物学原理的能力。在细胞周期中复制和转录之间的关系也将被阐明。我们还致力于为大规模分子生物学数据开发第一个数据驱动的预测张量计算模型。这些数据的结构比矩阵的结构高一个数量级，特别是当整合来自不同研究的数据时。将数据平铺成矩阵后，丢失了很多信息。我们将分析地研究几种可能的张量框架，并实现算法来计算和可视化它们。我们将把这些数学工具应用于研究癌症、细胞增殖和细胞周期的生物学数据。这个程序将导致新的见解到癌症，细胞增殖和细胞周期的生物程序之间的相互联系。我们的目标是在分子水平上更好地理解并最终控制生命过程。这些模型可能成为未来生物系统像今天的物理系统一样建模的基础。预测的调节机制可能是未来细胞分裂周期和癌症可以控制的基础。

DESCRIPTION (provided by applicant): Our research is motivated by recent advances in high-throughput technologies, such as DNA microarrays, which make it possible to record the complete genomic signals that guide the progression of cellular processes. Future predictive power and discovery in biology and medicine will come from the mathematical modeling of the rapidly growing number of these large-scale molecular biological datasets. To this end, we built the first data-driven predictive models for genomic data using frameworks from matrix computation. We illustrated thesse models in the analyses of, e.g., cell cycle expression data and transcription factors' and replication initiation proteins' DNA-binding data. The power of our models to predict previously unknown biological principles was demonstrated with a prediction of a novel mechanism of regulation that correlates replication initiation with cell cycle-regulated transcription in yeast. Now, we aim to validate experimentally this computational prediction by collecting and analyzing genome- wide expression under conditions that are thought to decouple replication from cell cycle transcription in eukaryotes. These experiments will test the ability of our mathematical models to correctly predict biological principles. The relation between replication and transcription during the cell cycle will also be illuminated. We also aim to develop the first data-driven predictive tensor computation models for large-scale molecular biological data. The structure of these data is of an order higher than that of a matrix, especially when integrating data from different studies. Flattened into a matrix much of the information in the data is lost. We will study analytically several possible tensor frameworks, and implement algorithms to compute and visualize them. We will apply these mathematical tools to biological data from studies of cancer, cellular proliferation and the cell cycle. This program will result with new insights into the interconnections among the biological programs of cancer, cellular proliferation and the cell cycle. Our goal is to enable better understanding and ultimately also control of life processes on the molecular level. These models may become the foundation of a future in which biological systems are modeled as physical systems are today. The predicted mechanism of regulation may be at the basis of a future where the cell division cycle and cancer can be controlled.