iMatch2: compound identification using retention index for analysis of gas chromatography-mass spectrometry data

Abstract

We developed a method, iMatch2, for compound identification using retention indices (RI) in NIST11 library. Three-way ANOVA test and Kruskal-Wallis test respectively demonstrate that column class and temperature program type defined by the NIST library are the most dominant factors affecting the magnitude of retention index while the retention index data type does not cause significant difference. The developed linear regression transformation for merging retention indices with different data types, but the same column class and temperature program type, reduces the standard deviation of retention index up to 8%, compared to the simple union approach used in the original iMatch. As for outlier detection methods to remove retention indices having large difference with the remaining data of the same compound, Tietjen-Moore test and generalized extreme studentized deviate test are the strictest methods, while methods such as Dixon's test, Thompson tau approach, and Grubbs' test are more conservative. To improve the accuracy of retention index window, a concept of compound specific retention index window is introduced for compounds with a large number of retention indices in the NIST11 library, while the retention index window is calculated from empirical distributions for the compounds with a small number of retention indices. Analysis of the experimental data of a mixture of compound standards and the metabolite extract from mouse liver show significant improvement of retention index quality in the NIST11 library and the new data analysis methods.

Keywords: Compound identification; GC–MS; Retention index; iMatch2.