药物分子设计中的大数据问题

发布时间：2019-03-05 16:59

【摘要】：药物创新领域的大数据主要来源于高通量实验、高效能模拟计算、信息化、科技出版物和专利文献4个方面.这些大数据使我们有可能在系统层面上看到药物分子与许多靶标相互作用的新现象、新规律,提高药物创新的效率,也带来新的挑战,如存储、标引/标注和质控、可视化、数据挖掘和计算复杂度等问题.这些问题可以通过在超算和云服务技术的支持下发展并行计算方法而逐渐得到解决.从离散、不完备且信噪比低的大数据中难以找到物质活性与结构之间的连续函数关系,贝叶斯学习机及其与支持向量机、决策树技术的组合是大数据挖掘的发展方向.大数据既是科学实验通量化和社会信息化的结果又是原因,正确解决大数据挖掘问题是提高药物创新效率的核心.
[Abstract]:Big data in the field of drug innovation mainly comes from four aspects: high-throughput experiment, high-efficiency simulation calculation, informatization, scientific and technological publications and patent literature. These big data have made it possible to see at the systematic level new phenomena, new laws, and new challenges such as storage, indexing / tagging and quality control, visualization of drug molecules interacting with many targets, and improving the efficiency of drug innovation. Data mining and computational complexity. These problems can be solved by developing parallel computing methods with the support of over-computing and cloud services. It is difficult to find the continuous function relationship between material activity and structure from big data, which is discrete, incomplete and low signal-to-noise ratio. The combination of Bayesian learning machine, support vector machine and decision tree technology is the development direction of big data mining. Big data is not only the result of scientific experiment and social informatization, but also the reason. It is the core to improve the efficiency of drug innovation to solve the problem of big data mining correctly.
【作者单位】：中山大学药学院药物分子设计与生物超算中心;合肥工业大学医学工程学院;
【基金】：国家自然科学基金(81173470) 国家高技术研究发展计划(2012AA020307) 广东省引进创新科研团队专项计划(2009010058) 广州超级计算应用研发与扶持专项(2012Y2-00048) 中央高校基本科研业务费专项(2013HGCH0015)资助
【分类号】：R91;TP311.13

【参考文献】