基于质谱图特征峰匹配的谱库预检索算法

Novel Algorithm for Spectral Library Pre-search Based on Spectral Feature Peak Matching

针对目前谱库预检索精度较低且剩余参考谱图依然较多的问题,提出了一种基于质谱图特征峰匹配的谱库预检索算法。将谱图中峰强度最高和质荷比最大的谱峰作为谱图特征峰,然后依次预检索了每个查询谱图和所有参考谱图对应的特征峰。结果表明,与相似度阈值分别为0.6、0.7和0.8的基于向量空间模型的谱库预检索算法相比,精度分别提高了15.01%、19.58%和25.87%,与二级谱库预检索算法相比,精度提高了1.77%,且该算法均能减少剩余参考谱图的数量,具有一定的优势。

A novel pre-search algorithm was proposed based on the spectral feature peak matching to address the existing problems of poor pre-retrieval precision and too many remaining spectra in mass spectral library search. In the newly-developed algorithm, the spectral peaks with the highest intensity and largest mass-to-charge ratio were defined as the characteristic peaks in the spectrum; then, each query spectrum was pre-retrieved by matching its characteristic peaks with those of all the reference spectra available. The results searched in both the novel and con- ventional algorithms were evaluated. The new algorithm outperforms TSLP algorithm with an accuracy improved by 1.77%. The new algorithm works better than the VSM-based Mass Spectral Library Pre-search algorithm, not only because the accuracy was improved by 15.01% ,19.58% and 25.87% ,respectively,with the thresholds of 0.6,0. 7 and 0.8, but also because the remaining spectra were significantly reduced.