Predicting Patient Outcomes from Clinical and Genome-Wide Data

描述(由申请人提供)： 临床分型和预测是临床护理的关键组成部分。即使在分类和预测性能方面略有改进，也可能在改善患者结局和降低医疗成本方面产生重大的医疗后果。这项建议描述了一种新的、高效的贝叶斯机器学习方法，用于执行临床预测。该方法旨在使用临床和全基因组数据来预测患者结果。 该方法的性能将使用弗雷明翰心脏研究的现有临床和全基因组数据进行评估，这些数据将通过国家生物技术信息中心的共享资源提供给合格的研究人员。在个体中预测的结果包括主要心血管事件的发生和各种原因造成的死亡。这项研究将评估单独使用传统临床数据、单独使用全基因组数据以及同时使用两种类型的数据时，使用新的贝叶斯方法进行预测的效果。该方法的性能也将与现有模型和方法进行比较。 要检验的主要假设是，拟议的机器学习方法将是对现有方法的进步，因为使用传统临床数据和全基因组数据的组合应用它在计算上是可行的，并且它将产生比现有预测模型和方法更好的预测性能。如果被证明是这样的话，这种新方法有望在未来的电子患者护理系统中提供实质性的好处，包括应用于基于计算机的决策支持，这些决策支持既可以获得传统的临床数据，也可以获得基因组范围的数据。

DESCRIPTION (provided by applicant): Clinical classification and prediction are key components of clinical care. Even modest improvements in classification and predictive performance may have significant healthcare consequences in terms of improved patient outcomes and reduced healthcare costs. This proposal describes a new, efficient Bayesian machine-learning method for performing clinical predictions. The method is designed to use both clinical and genome-wide data in predicting patient outcomes. The method's performance will be evaluated using existing clinical and genome-wide data from the Framingham Heart Study that is being made available to qualified researchers through the SHARe resource of the National Center for Biotechnology Information. The outcomes to be predicted in individuals include the onset of major cardiovascular events and death from all causes. The study will evaluate how well these predictions can be made with a new Bayesian method when using traditional clinical data alone, genome-wide data alone, and both types of data together. The method's performance will also be compared that of existing models and methods. The main hypothesis to be tested is that the proposed machine-learning method will be an advancement over existing methods in that it will be computationally feasible to apply it using a combination of traditional clinical data and genome-wide data, and it will yield better predictive performance than do existing predictive models and methods. If shown to be so, this new method is anticipated to provide substantial benefit in future electronic patient-care systems, including applications to computer-based decision support that have available both traditional clinical data and genome-wide data.