基于Jaccard系数提高核磁共振波谱代谢物数据库匹配的精度

发布时间：2018-03-14 14:27

本文选题：代谢组学　切入点：核磁共振　出处：《化学研究与应用》2017年12期 　论文类型：期刊论文

【摘要】：NMR代谢组学检测完成后,人们通常基于化学位移值在人类代谢组学数据库(human metabolome database,HMDB)上进行手动代谢物匹配,然而该方法对代谢物的鉴定较为粗糙,准确度不高。本研究试图基于建立一种更加合理,且能够自动寻峰并根据数据库匹配代谢物方法。通过分析HMDB的峰匹配方法,提出了基于Jaccard系数和匹配率(匹配的峰数目/总峰数)的新方法,基于MATLAB编程实现,然后比较HMDB中1D NMR search和本方法对于同一段随机化学位移列表的匹配结果。分析结果显示,对于同一随机化学位移列表,HMDB的匹配结果中排在前20位的物质峰数目超过16的占60%,说明其匹配方法偏向于峰数目较多的物质;HMDB用于峰匹配排序的评分与峰匹配率有明显区别,而本方法匹配评分与匹配率较为接近;且HMDB匹配结果排在第10位的物质与该随机序列没有可匹配的化学位移值。本文对于HMDB峰匹配算法存在的不足进行了改进,并发现基于Jaccard分数的匹配算法能够提高根据代谢物数据库进行NMR代谢物鉴定的精度。
[Abstract]:After the NMR metabonomics test is completed, manual metabolite matching is usually performed on the human metabolome database HMDBs based on the chemical shift values. However, the identification of metabolites by this method is rather rough. This study is based on the establishment of a more reasonable and automatic peak searching method and matching metabolites according to the database. By analyzing the peak matching method of HMDB, A new method based on Jaccard coefficient and matching rate (matching peak number / total peak number) is proposed. It is realized by MATLAB programming. Then, the matching results of 1D NMR search and this method for the same random chemical shift list in HMDB are compared. For the same random chemical shift list of HMDB matching results, the number of the top 20 peaks is more than 16, which indicates that the matching method is inclined to the substances with more peaks than HMDB, and there is a significant difference between the score and the peak matching rate of the HMDB for the matching ranking of the peaks. The matching score of this method is close to the matching rate, and the material in the tenth place of HMDB matching results has no matched chemical shift value with the random sequence. In this paper, the shortcomings of the HMDB peak matching algorithm are improved. It is found that the matching algorithm based on Jaccard fraction can improve the accuracy of NMR metabolites identification based on metabolites database.
【作者单位】：电子科技大学医学院;四川省医学科学院·四川省人民医院创伤代谢组多学科实验室;四川大学华西基础医学与法医学院组织胚胎学教研室;
【基金】：四川省科技厅科技支撑计划项目(2014FZ0125、2015SZ0110)资助
【分类号】：O657.2;R313

【相似文献】