BIGDATA: Mid-Scale DCM: DA: ESCE: Discovering Molecular Processes

描述（由申请人提供）：癌症是一种基因组疾病，由人的DNA破坏引起。价格的数量级下降和测序通量的增加使得能够对数百个基因组进行测序。这将启动一个新的阶段BF "精准医学"，其中分子标记物可以指导为患者量身定制的治疗。基因组学革命现在正在系统地表征大型队列（> 300名患者）中每个肿瘤的每个体细胞变化。尽管取得了一些成功，但基于分子特征预测癌症结果仍然是一个重大挑战。这项建议旨在消除阻碍进展的几个主要障碍。首先，原始序列数据不是特别适合用于开发预测模型。因此，基因和途径水平的证据将来自CGHub，以显着提高生物医学发现信息的实用性。这些信息将被收集在Facebook等社交图谱技术框架中，以扩展到数十亿个相互关联的对象，称为生物医学证据图（BMEG）。其次，数据集太大，在互联网上移动是不切实际的。因此，将创建一个环境，研究人员可以在其中将他们的代码移动到BMEG中的大量数据。第三，预测挑战将基于癌症基因组学数据集和患者结果。虽然在预测结果方面取得了一些成功，但目前的方法在应用于看不见的数据时，存在可重复性和鲁棒性问题。该活动将覆盖广泛的算法开发人员社区，促进生物信息学代码的透明度和共享，并创造强大的网络效应，以众包开发生物发现的最佳模型。构建的系统将专注于癌症结果的研究，但整个管道将用于任何数量的基于基因组的项目，包括研究任何数量的疾病，干细胞特性，模式生物和全基因组关联研究。相关性（参见说明）：虽然我们积累了大量关于癌细胞的信息，但我们仍在黑暗中寻找预测治疗策略的线索。加速计算发现至关重要。BMEG的创建将促进社区参与，以揭示新的关系，阐明肿瘤发生的新基础生物学和治疗这种疾病的治疗方向。

DESCRIPTION (provided by applicant): Cancer is a disease of the genome, caused by disruptions in a person's DNA. Orders of magnitude decreases in price and increases in sequencing throughput enabled sequencing of hundreds of genomes. This will launch a new phase bf "Precision Medicine," where molecular markers can guide therapies tailored to patients. The genomics revolution is now systematically characterizing every somatic change in every tumor for large cohorts (>300 patients). Despite some successes, predicting cancer outcomes based on molecular signatures remains a major challenge. This proposal aims at obliterating several key roadblocks stymieing progress. First, the raw sequence data is not particularly well-suited for use in developing predictive models. Therefore, gene- and pathway-level evidence will be derived from CGHub to significantly increase the utility of the information for biomedical discovery. The information will be collected in a Social Graph technology framework like Facebook to scale to billions of interconnected objects called the Biomedical Evidence Graph (BMEG). Second, the datasets are so large they are impractical to move around on the internet. Thus, an environment will be created within which researchers can move their code to the vast amounts of data within the BMEG. Third, prediction challenges will be created based on cancer genomics datasets and patient outcomes. While there have been a few successes in predicting outcomes, current approaches suffer from reproducibility and robustness when applied to unseen data. This activity will reach a broad community of algorithm developers, promote transparency and sharing of bioinformatics code, and create a strong network effect to crowd-source the development of the best models for biological discovery. The system constructed will be focused around the investigation of cancer outcomes but the entire pipeline will be of general utility for any number of genome-based projects including investigating any number of disease, stem cell properties, model organisms, and genome-wide association studies. RELEVANCE (See instructions): While we are accumulating vast amounts of information on cancer cells, we are still searching in the dark for clues about predicting treatment strategies. It is of paramount importance to accelerate computational discovery. The creation of the BMEG will catalyze community participation to uncover novel relationships to elucidate new fundamental biology on oncogenesis and therapeutic directions for treating this disease.