Multi-Tensor Decompositions for Personalized Cancer Diagnostics and Prognostics

 描述(申请人提供)：复发性DNA拷贝数改变(CNA)100年来一直被认为是癌症的标志，但这些改变对肿瘤的发病机制以及患者的诊断、预后和治疗仍知之甚少。这是尽管记录单个疾病的不同方面的大规模多维数据集的数量不断增加，例如在癌症基因组图谱(TCGA)中，并且由于对数学框架的基本需求，所述数学框架可以从以匹配列的多个张量布置的这样的多个数据集创建一个相干模型，所述匹配列例如患者、平台和组织，但是独立的行，例如探针。例如，我们最近对患者匹配的胶质母细胞瘤(GBM)脑瘤和来自TCGA(排列在两个矩阵中，列相匹配但独立的行)的正常血液基因组图谱进行的比较建模(通过使用数据驱动的双矩阵谱分解)揭示了一种以前未知的肿瘤排他性CNA的全球模式，该模式与GBM存活和化疗反应相关，并且可能是因果相关的。自2008年以来，这些数据一直是公开的，但直到我们在2012年应用我们的比较模型之前，这个签名一直不为人所知。生存分析显示，并经计算验证，该签名比年龄表现更好，并且在统计上独立于年龄，年龄是50年来GBM存活的最佳指标，以及现有的GBM病理实验室测试。基于这一签名的一项新的GBM测试正在等待犹他州大学病理学系的非营利性参考实验室--相关地区和大学病理学家(ARUP)实验室，Inc.的实验重新验证。在这个NCI U01项目中，我们来自生物工程、数学和病理学系、科学计算和成像(SCI)研究所以及犹他大学亨茨曼癌症研究所(HCI)的多学科研究团队旨在(I)定义和研究数据驱动的多张量光谱分解的特性；(Ii)使用这些分解来模拟患者、平台和组织匹配但不依赖于探针的TCGA基因组图谱，并获得对较低级别星形细胞(LGA)脑癌、卵巢浆液性囊腺癌(OV)和肺鳞癌的基因-表型关系的生物学和医学见解；以及(Iii)通过实验测试现有GBM模型以及使用犹他州样本的新型LGA和OV模型的计算预测，使这些见解能够转化为病理学实验室测试。最终，这个项目将使医生更接近于有一天能够预测和控制细胞分裂和癌症的进展，就像NASA工程师今天绘制航天器轨迹一样容易。

 DESCRIPTION (provided by applicant): Recurring DNA copy-number alterations (CNAs) have been recognized as a hallmark of cancer for >100 years, yet what these alterations imply about a tumor's pathogenesis and a patient's diagnosis, prognosis, and treatment remains poorly understood. This is despite the growing number of large-scale multidimensional datasets recording different aspects of a single disease, e.g., in the Cancer Genome Atlas (TCGA), and due to a fundamental need for mathematical frameworks that can create one coherent model from such multiple datasets arranged in multiple tensors of matched columns, e.g., patients, platforms, and tissues, but independent rows, e.g., probes. For example, our recent comparative modeling (by using a data-driven two-matrix spectral decomposition) of patient-matched glioblastoma (GBM) brain tumor and normal blood genomic profiles from TCGA (arranged in two matrices, of matched columns but independent rows) uncovered a previously unknown global pattern of tumor-exclusive CNAs that is correlated with, and possibly causally related to, GBM survival and response to chemotherapy. The data had been publicly available since 2008, but this signature remained unknown until we applied our comparative modeling in 2012. Survival analyses showed, and computationally validated, that the signature performs better than, and is statistically independent of, age, the best indicator of GBM survival for >50 years, and existing GBM pathology laboratory tests. A new test for GBM based upon this signature is pending an experimental re-validation at the Associated Regional and University Pathologists (ARUP) Laboratories, Inc., a nonprofit reference laboratory of the Department of Pathology at the University of Utah. In this NCI U01 project, our multidisciplinary team of researchers from the Departments of Bioengineering, Mathematics, and Pathology, the Scientific Computing and Imaging (SCI) Institute, and the Huntsman Cancer Institute (HCI) at the University of Utah, aims to (i) define, and study the properties of data- driven multi-tensor spectral decompositions; (ii) use these to model patient-, platform-, and tissue-matched but probe-independent TCGA genomic profiles, and gain biological and medical insights into the genotype- phenotype relations in lower-grade astrocytoma (LGA) brain cancer, ovarian serous cystadenocarcinoma (OV), and lung squamous cell carcinoma; and (iii) enable translation of these insights into pathology laboratory tests, by experimentally testing the computational predictions of the existing GBM model, as well as the novel LGA and OV models by using Utah samples. Ultimately, this project will bring physicians a step closer to one day being able to predict and control the progression of cell division and cancer as readily as NASA engineers plot the trajectories of spacecraft today.