Statistical and Machine Learning Methods to Address Biomedical Challenges for Integrating Multi-view Data

项目摘要 许多疾病都是复杂的、异质性的，会影响身体的多个器官，并取决于 多种因素之间的相互作用，包括遗传、细胞、分子和环境因素。因此，它是 不足为奇的是，许多复杂疾病的发病机制仍然难以捉摸，治疗靶点也缺乏。 关注少数分子(例如，基因或代谢物)或单一类型的 数据(例如，临床或遗传)不能解决这种复杂性和异质性。综合生物学或系统生物学 可以使用方法和网络分析来利用来自多个源的数据的优势(例如， 基因组学、代谢组学、流行病学、临床数据)，以实现对复杂疾病病理生物学的新见解 疾病。最近的技术进步使得产生大量不同但相关的数据成为可能 具有丰富的信息，为了解复杂的生物过程提供了极好的机会 疾病和变革医学，但同时提出了重大的fi无法分析的挑战，包括如何 从数以万计的数据点中有效地合成信息以识别重要的生物标志物 具有作为治疗靶点的潜力。为了缓解这一问题，我们将开发和应用一套新颖、健壮、 以及强大的统计和机器学习方法，用于综合和解释横断面和 来自多个来源的纵向数据。这些模型还将用于fiNe亚群患者 根据来自不同来源的数据，有不同的预后或需要不同的治疗方法。此外， 我们将利用网络理论的最新进展来模拟分子中复杂的多边关系 来自多个来源的数据。建议的方法将应用于几个公开可用的数据集和队列 以确保我们可以将我们的工作推广到其他数据集和队列，从而增加长期影响 我们的研究成果。拟议的研究还将对统计和机器学习算法做出有价值的贡献 这将广泛适用于来自多个来源和多个队列的数据，并将提供给 免费向公众开放。

Project Summary Many diseases are complex, heterogeneous, conditions that affect multiple organs in the body and depend on the interplay between several factors that include genetic, cellular, molecular, and environmental factors. It is therefore not surprising that the pathogenesis of many complex diseases remain elusive, and therapeutic targets are lacking. The traditional approach that focus on a small number of molecules (e.g., genes or metabolites) or a single type of data (e.g., clinical or genetic) cannot address this complexity and heterogeneity. Integrative or systems biology approaches and network analysis can be used to leverage the strengths of data from multiple sources (e.g., genomics, metabolomics, epidemiology, clinical data) to achieve new insights into the pathobiology of complex diseases. Recent technological advances have enabled the production of vast amounts of diverse but related data with rich information that offer remarkable opportunities to understand biological processes involved in complex diseases and to transform medicine, yet at the same time present significant analytical challenges including how to effectively synthesize information from the tens of thousands of data points to identify important biomarkers with potential to serve as therapeutic targets. To alleviate this, we will develop and apply a suite of novel, robust, and powerful statistical and machine learning methods for the integration and interpretation of cross-sectional and longitudinal data from multiple sources. These models will also be used to define subpopulations of patients who have different prognoses or require different therapeutic approaches based on data from different sources. Further, we will make use of recent advances in network theory to model the complex multilateral relationships in molecular data from multiple sources. The proposed methods will be applied to several publicly available datasets and cohorts to ensure that we can generalize our work to other datasets and cohorts and thus increase the long-term impact of our research. The proposed research will also contribute valuable statistical and machine learning algorithms that will be broadly applicable to data from multiple sources and multiple cohorts and will be made available to the public free of charge.