Wathieu Henri, Issa Naiem T, Mohandoss Manisha, Byers Stephen W, Dakshanamurthy Sivanesan
Department of Oncology, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington DC 20057. United States.
Department of Biochemistry & Molecular Biology, Georgetown University, Washington DC 20057. United States.
Comb Chem High Throughput Screen. 2017;20(3):193-207. doi: 10.2174/1386207319666161214111254.
Cancer-associated metabolites result from cell-wide mechanisms of dysregulation. The field of metabolomics has sought to identify these aberrant metabolites as disease biomarkers, clues to understanding disease mechanisms, or even as therapeutic agents.
This study was undertaken to reliably predict metabolites associated with colorectal, esophageal, and prostate cancers. Metabolite and disease biological action networks were compared in a computational platform called MSD-MAP (Multi Scale Disease-Metabolite Association Platform).
Using differential gene expression analysis with patient-based RNAseq data from The Cancer Genome Atlas, genes up- or down-regulated in cancer compared to normal tissue were identified. Relational databases were used to map biological entities including pathways, functions, and interacting proteins, to those differential disease genes. Similar relational maps were built for metabolites, stemming from known and in silico predicted metabolite-protein associations. The hypergeometric test was used to find statistically significant relationships between disease and metabolite biological signatures at each tier, and metabolites were assessed for multi-scale association with each cancer. Metabolite networks were also directly associated with various other diseases using a disease functional perturbation database.
Our platform recapitulated metabolite-disease links that have been empirically verified in the scientific literature, with network-based mapping of jointly-associated biological activity also matching known disease mechanisms. This was true for colorectal, esophageal, and prostate cancers, using metabolite action networks stemming from both predicted and known functional protein associations.
By employing systems biology concepts, MSD-MAP reliably predicted known cancermetabolite links, and may serve as a predictive tool to streamline conventional metabolomic profiling methodologies.
癌症相关代谢物源自细胞整体的失调机制。代谢组学领域一直致力于将这些异常代谢物鉴定为疾病生物标志物、理解疾病机制的线索，甚至作为治疗药物。
本研究旨在可靠地预测与结直肠癌、食管癌和前列腺癌相关的代谢物。在一个名为MSD-MAP（多尺度疾病-代谢物关联平台）的计算平台上比较代谢物和疾病生物作用网络。
利用来自癌症基因组图谱的基于患者的RNA测序数据进行差异基因表达分析，确定与正常组织相比在癌症中上调或下调的基因。使用关系数据库将包括通路、功能和相互作用蛋白在内的生物实体映射到那些差异疾病基因上。基于已知的和计算机预测的代谢物-蛋白质关联，为代谢物构建类似的关系图谱。使用超几何检验来发现各层级疾病和代谢物生物特征之间的统计学显著关系，并评估代谢物与每种癌症的多尺度关联。还使用疾病功能扰动数据库将代谢物网络直接与各种其他疾病相关联。
我们的平台重现了科学文献中已通过实验验证的代谢物-疾病联系，基于网络的联合相关生物活性映射也与已知疾病机制相符。对于结直肠癌、食管癌和前列腺癌，使用源自预测和已知功能蛋白关联的代谢物作用网络时都是如此。
通过运用系统生物学概念，MSD-MAP可靠地预测了已知的癌症-代谢物联系，并可作为一种预测工具来简化传统的代谢组学分析方法。
