Causal Discovery Algorithms for Translational Research with High-Throughput Data

项目摘要 面向高吞吐量数据翻译研究的因果发现算法 该项目的长期目标是为生物医学界提供下一代因果 算法，以促进发现疾病的分子途径和病因以及预测 生物标志物和高通量数据的分子特征。这些知识和方法， 对早期和更准确的诊断和预后，个性化医疗， 合理的药物设计 如果成功的话，所提出的研究将具有重要和广泛的方法论和实践意义。 影响跨越生物医学的几个领域，主要关注和直接利益， 高通量诊断和个性化医疗。它将大大改善 计算方法和更深层次的理论理解， 签名和了解疾病的机制和新药的伴随线索。它将 提供证据证明新的因果关系方法在其他类型的数据中的适用性。它将产生 人类肺癌的具体途径的见解。它将加深我们的理解和解决方案 经济学数据中的罗生门效应。拟议的研究也将阐明业务 稳定性启发式的价值。最后，这项研究将吸引国际研究界的参与， 解决与高通量和其他相关的开放计算因果发现问题， 生物医学数据。 目标1.评估和表征生物标志物的几种新型因果算法 选择，分子签名创建和反向网络工程，使用真实的，模拟的， 再模拟和实验数据集。研究方法的通用性， 适用于非组学数据集。 目标2。评估和表征，新颖和最先进的因果算法， 最先进的非因果和准因果算法。 目标3。系统地研究罗生门效应，因为它适用于生物标志物， 签名多重性。 目标4。系统地研究应用稳定性启发式的效用， 因果发现 目标5。推导出肺癌的新生物标志物、途径和假说。 目标6。通过国际因果发现竞赛得出新颖的解决方案。 目标7。传播调查结果。

Project Summary Causal Discovery Algorithms for Translational Research with High-Throughput Data The long-term goal of this project is to provide to the biomedical community next-generation causal algorithms to facilitate discovery of disease molecular pathways and causative as well as predictive biomarkers and molecular signatures from high-throughput data. Such knowledge and methods are necessary toward earlier and more accurate diagnosis and prognosis, personalized medicine, and rational drug design. If successful, the proposed research will have significant and wide methodological and practical implications spanning several areas of biomedicine with a primary focus and immediate benefits in high-throughput diagnostics and personalized medicine. It will provide significantly improved computational methods and deeper theoretical understanding related to producing molecular signatures and understanding mechanisms of disease and concomitant leads for new drugs. It will provide evidence about applicability of novel causal methods in other types of data. It will generate insights in specific pathways of lung cancer in humans. It will deepen our understanding and solutions to the Rashomon effect in ¿omics¿ data. The proposed research will also shed light on the operational value of the stability heuristic. Finally the research will engage the international research community to address open computational causal discovery problems relevant to high-throughput and other biomedical data. ¿ Aim 1. Evaluate and characterize several novel causal algorithms for biomarker selection, molecular signature creation and reverse network engineering using real, simulated, resimulated, and experimental datasets. Study generality of the methods by means of applicability to non-¿omics¿ datasets. ¿ Aim 2. Evaluate and characterize, novel and state of the art causal algorithms against state-of-the-art non-causal and quasi-causal algorithms. ¿ Aim 3. Systematically investigate the Rashomon effect as it applies to biomarker and signature multiplicity. ¿ Aim 4. Systematically investigate the utility of applying the stability heuristic for causal discovery. ¿ Aim 5. Derive novel biomarkers, pathways and hypotheses for lung cancer. ¿ Aim 6. Induce novel solutions through an international causal discovery competition. ¿ Aim 7. Disseminate findings.