Gene expression programs of lactic acidosis in human cancers

描述（由申请人提供）：人类癌症具有极强的异质性。综合癌症系统生物学必须发展出更全面、更深刻的方法来探究异质性的各种来源，并在针对单个肿瘤的精细预测模型中合理地表达它们复杂的相互作用。癌症的一个重要特征是存在各种微环境应激，如缺氧、高乳酸和细胞外酸中毒。这些压力触发癌细胞的表型变化，并直接调节生理、代谢和最终的临床表型。虽然这些特征的重要性是公认的，他们通常不纳入临床决策，因为显著的概念和后勤困难。在我们最初的R01拨款中，我们提出通过基于微阵列的“基因组基因签名”来整合乳酸酸中毒的影响。任何一个“特征”都是由一组基因组成的，这些基因的表达水平在体外被特定的应激改变，因此可能表明微环境对体内肿瘤状态的影响。这些基因特征是实验细胞培养物和患者肿瘤共享的“常见表型”，允许信息在不同背景之间相互流动：从体外机制实验研究到癌症患者的体内分子和临床表型，然后返回。然而，在癌症和其他情况下，体外细胞操作不能完全反映人类癌症的复杂性和多样性。因此，我们建议包括额外的co-PI， Lucas博士，以应用稀疏统计因子模型，并使用人类癌症的表达数据作为“分子棱镜”，人们可以将体外特征分解和细化为体内成分“因素”或“子特征”，它们在人类癌症中以复杂的方式共变和相互作用。这些“子特征”因素保留了它们与原始特征的关系，但代表了由初始特征所代表的生物过程中不同的、相互作用的组成部分。这些解剖的体内特征的数值总结使我们能够研究这些反应是如何单独或集体地与癌症中的致癌信号事件、DNA拷贝变化、其他基因组数据以及它们与临床结果的关系联系起来的。许多子特征表达水平的变化与DNA拷贝数的变化直接相关。我们将研究这些在DNA水平上的变化如何影响癌细胞对缺氧和乳酸酸中毒的反应。拟议的实验将扩大我们最初的资助范围，包括全面、定量地描述乳酸酸中毒和缺氧反应与基因组和遗传因素相互作用的复杂性，并将这些相互作用与影响患者预后的复杂癌症表型相关联。

DESCRIPTION (provided by applicant): Human cancers are extremely heterogeneous. Integrative cancer systems biology must develop more comprehensive and incisive methods to interrogate various sources of heterogeneity, and to rationally represent their complex interactions in refined predictive models for individual tumors. One important feature of cancers is the presence of various micro- environmental stresses, such as oxygen depletion, high lactate and extracellular acidosis. These stresses trigger phenotypic changes in cancer cells and directly modulate physiological, metabolic and ultimately clinical phenotypes. Although the importance of these features is well recognized, they are not usually incorporated in clinical decision making because of the significant conceptual and logistic difficulties. In our original R01 grant, we proposed to integrate the influence of lactic acidosis through "genomic gene signatures" based on microarrays. Any one "signature" consists of a set of genes whose expression levels are altered by specific stress in vitro and may therefore be indicative of the micro-environmental effects on the tumor states in vivo. These gene signatures are "common phenotypes" shared by experimental cell cultures and patient tumors to permit the reciprocal flow of information between the contexts: from in vitro mechanistic experimental studies to in vivo molecular and clinical phenotypes of cancer patients, and back. However, in cancer as in other contexts, in vitro cell manipulations cannot fully reflect the complexity and variation seen in human cancers. Therefore, we propose to include additional co-PI, Dr. Lucas to apply sparse statistical factor models and use the expression data of human cancers as "molecular prisms" with which one may dissect and refine the in vitro signatures into in vivo component "factors" or "sub-signatures" which co-vary and interact in complex ways among human cancers. These "sub-signature" factors retain their relationship to the original signature but represent distinct, interacting components of the biological processes represented by the initial signature. Numerical summaries of these dissected in vivo signatures allow us to investigate how these responses are, individually or collectively, linked to oncogenic signaling events in cancers, to DNA copy changes, to other genomic data and to their relationship with clinical outcomes. Variation in the expression levels of many sub-signatures are directly related to variations in the DNA copy number. We will investigate how these changes at the DNA level in cancer cell lines will affect their response to hypoxia and lactic acidosis. The proposed experiments will broaden the scope of our original grant to include comprehensive, quantitative descriptions of the complexity of the interactions of the lactic acidosis and hypoxia responses with genomic and genetic factors, and correlate these interactions with the complex cancer phenotypes that impact patient outcomes. PUBLIC HEALTH RELEVANCE: In this application responsive to the Notice NOT-OD-09-058 titled: NIH Announces the Availability of Recovery Act Funds for Competitive Revision Applications, we propose to apply sparse statistical factor models and use the expression data of human cancers as "molecular prisms" with which one may dissect and refine the in vitro signatures into in vivo component "factors" or "sub-signatures" which co-vary and interact in complex ways among human cancers. These "sub-signature" factors retain their relationship to the original signature but represent distinct, interacting components of the biological processes represented by the initial signature. Numerical summaries of these dissected in vivo signatures allow us to investigate how these responses are, individually or collectively, linked to oncogenic signaling events in cancers, to DNA copy changes, to other genomic data and to their relationship with clinical outcomes.