Building Novel Predictive Networks for high-throughput, in-silico Key Driver Prioritization to Enhance Drug Target Discovery in AMP-AD and M2OVE-AD

项目摘要 我们响应RFA（AG-17-054）的目标是：1）应用新的计算方法，即顶部- 向下和自底向上的预测网络（简称预测网络）到现有的丰富的数据集， AMP-AD和M2 OVE-AD联盟发现可以指导治疗开发的新靶点; 2） 使用人类细胞模型进行新靶标的实验验证，并生成RNA-seq 数据，以系统地验证计算机预测; 3）整合新的RNA-seq数据，以进一步提高 预测网络建立在现有的AMP-AD数据和提高目标的质量。阿尔茨海默 痴呆症是痴呆症最常见的形式，估计影响全球3600万人。这个数字 预计到2050年将增加到1.15亿，除非开发出有效的治疗方法。AMP-AD目标 发现和临床前验证以及M2 OVE-AD项目是大规模的开放科学联盟，旨在 在建立一个预测性的，多尺度的AD模型，更好地反映其异质性，并在发现 通过综合的、数据驱动的方法来开发下一代治疗靶点。虽然四个多- AMP-AD的机构、多学科团队参与了前沿和不可知的分析工作， 重建AD的基因、蛋白质和代谢产物的分子网络，并发现新的治疗方法。 通过多个模型生物体评价的靶点，在所有模型生物体中提出的药物靶点的重现性 团队是低的（1个重复的80个目标，提出了在9月。2016年），由于 数据类型中存在的疾病和差异以及用于生成和分析 数据这种变化削弱了对每个候选目标的信心，分散了 实验验证目前的关键驱动筛选缺乏高通量、计算机模拟筛选组件， 通过预测下游分子表型直接评估靶标的功效， 扰动目标发现和验证之间的这种差距进一步降低了总体成功率， 药物开发的有效性和效率。在本提案中，我们将应用预测网络管道， 与AMP-AD中的所有团队合作，对AMP-AD中的现有数据进行计算机筛选， M2 OVE-AD建立因果和预测网络，并确定和优先考虑关键驱动因素，这将是 通过实验验证进行评价。通过实验验证生成的RNA-seq数据将用于 系统评价计算机模拟表型预测，以确定拟定治疗方案的置信度 配方，并与现有的预测网络相结合，以加强药物靶点的发现。

Project Summary We respond to the RFA (AG-17-054) with the goals of 1) applying new computational approaches, i.e. top- down and bottom-up predictive network (predictive network for short) to the existing rich datasets generated by the AMP-AD and M2OVE-AD consortia to discover novel targets that can guide therapy development; 2) performing experimental validation of the novel targets using human cellular models and generating RNA-seq data to systematically validate in-silico prediction; 3) integrating the new RNA-seq data to further improve the predictive network built on existing AMP-AD data and to enhance the quality of the targets. Alzheimer's disease is the most common form of Dementia estimated to affect 36 million people worldwide. This number is expected to rise to 115 million by 2050 unless an effective therapeutic is developed. The AMP-AD Target Discovery and Preclinical Validation and M2OVE-AD programs are large-scale, open science consortia aimed at building a predictive, multi-scale model of AD that better reflects its heterogeneity and at discovering the next-generation therapeutic targets through integrative, data-driven approaches. While the four multi- institutional, multidisciplinary teams in AMP-AD engaged cutting-edge and agnostic analysis efforts to reconstruct the molecular network of the gene, protein, and metabolite in AD and to discover novel therapeutic targets evaluated by multiple model organisms, the reproducibility of the drug targets proposed across all teams is low (1 replicate out of total 80 targets proposed in Sept. 2016) given the considerable complexity of the disease and differences that exist in the data types and the approaches used to generate and analyze the data. This variability undermined the confidence of each candidate target and defocused the efforts of experiment validation. Current key driver screening lacks a high-throughput, in-silico screening component to directly evaluate the efficacy of a target by predicting the downstream molecular phenotype given its perturbation. This gap between target discovery and validation further reduces the overall rate of success, efficacy and efficiency of drug development. In this proposal, we will apply the predictive network pipeline with the in-silico screening component to existing data in AMP-AD in collaboration with all teams in AMP-AD and M2OVE-AD to build causal and predictive networks and to identify and prioritize key drivers, which will be evaluated by experiment validation. The RNA-seq data generated by experiment validation will be used to systematically evaluate the in-silico phenotypic prediction to determine the confidence of proposed therapeutic recipes and to be integrated with the existing predictive networks to enhance drug target discovery.