人工智能培训

搜索

人工智能论文:使用交互增强个体特征预测能力(Empowering individual trait prediction using interactions)

[复制链接]
imissa 发表于 2019-1-28 12:02:31 | 显示全部楼层 |阅读模式
imissa 2019-1-28 12:02:31 454 0 显示全部楼层
人工智能论文:使用交互增强个体特征预测能力(Empowering individual trait prediction using interactions)精确医学的一个组成部分是构建具有尽可能高的预测能力的预测模型,例如,实现个人风险预测。在遗传流行病学中,复杂疾病具有多基因基础,并且通常的假设是生物学和遗传学特征通过相互作用影响所考虑的结果。在组学数据的情况下,诸如广义线性模型之类的标准方法的使用可以是次优选地机器学习方法对于进行单独预测是有吸引力的。然而,这些算法中的大多数主要关注于数据集中的单个特征的主要或边际效应。另一方面,相互作用特征的检测是遗传性流行病学领域的研究活跃领域。检测交互特征的一大类算法基于多因素降维(MDR)。在这里,我们扩展基于模型的MDR(MB-MDR),这是原始MDR算法的强大扩展,以实现交互授权的个体预测。使用综合模拟研究,我们表明我们的新算法可以比其他两种最先进的算法,即随机森林和弹性网更有效地使用隐藏在交互中的信息,并且如果存在交互,则明显优于这些。如果不存在相互作用,则这些算法的性能是可比较的。此外,我们通过比较三种算法在类风湿性关节炎病例和健康对照数据集上的表现,表明我们的新算法适用于实际数据。由于我们的新算法不仅适用于生物/遗传数据,而且适用于具有离散特征的所有数据集,它也可能在其他应用中具有实际意义,并且我们将我们的方法作为R包提供。
One component of precision medicine is to construct prediction models withtheir predictive ability as high as possible, e.g.to enable individual riskprediction.In genetic epidemiology, complex diseases have a polygenic basisand a common assumption is that biological and genetic features affect theoutcome under consideration via interactions.In the case of omics data, theuse of standard approaches such as generalized linear models may be suboptimaland machine learning methods are appealing to make individual predictions.However, most of these algorithms focus mostly on main or marginal effects ofthe single features in a dataset.On the other hand, the detection ofinteracting features is an active area of research in the realm of geneticepidemiology.One big class of algorithms to detect interacting features isbased on the multifactor dimensionality reduction (MDR).Here, we extend themodel-based MDR (MB-MDR), a powerful extension of the original MDR algorithm,to enable interaction empowered individual prediction.Using a comprehensivesimulation study we show that our new algorithm can use information hidden ininteractions more efficiently than two other state-of-the-art algorithms,namely the Random Forest and Elastic Net, and clearly outperforms these ifinteractions are present.The performance of these algorithms is comparable ifno interactions are present.Further, we show that our new algorithm isapplicable to real data by comparing the performance of the three algorithms ona dataset of rheumatoid arthritis cases and healthy controls.As our newalgorithm is not only applicable to biological/genetic data but to all datasetswith discrete features, it may have practical implications in otherapplications as well, and we made our method available as an R package.人工智能论文:使用交互增强个体特征预测能力(Empowering individual trait prediction using interactions) WII9N1EFxUE5Zn29.jpg
URL地址:https://arxiv.org/abs/1901.08814     ----pdf下载地址:https://arxiv.org/pdf/1901.08814    ----人工智能论文:使用交互增强个体特征预测能力(Empowering individual trait prediction using interactions)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

imissa当前离线
新手上路

查看:454 | 回复:0

快速回复 返回顶部 返回列表