Optimization of the automatic annotation of triglycerides in mass spectrometry retention chromatography spectra in R

Abstract

The development of high-performance mass spectrometry (MS) techniques in combination with high performance liquid chromatography (HPLC) has become indispensable in lipidomics studies. As lipidomics produces a vast amount of data, robust bioinformatics is needed for analysis. After data processing and analysis, lipid annotation can be performed. However, lipid annotation methods have not been optimized yet and are dependent on the manual drawing of polygons around lipid classes. In this study an attempt was made to find an alternative approach for the annotation of triglycerides (TGs) in mass spectrometry (MS) chromatography analyses that was not dependent on the manual drawing of polygons. As non-polar lipids, such as triglycerides (TGs), diacylglycerols (DGs), and ceramides, tend to show a linear pattern in the graphical representations of RT-MS spectra, the use of linear regression models was proposed. The regression models were developed with the dataset deduced from the manual polygonal drawing method and compared to an alternative algorithm, that was not dependent on the manual polygonal drawing method. Several linear regression models were developed to investigate the correlation between mass and retention time: a univariate, a two-variate, and a multivariate regression model were created. Strong correlations (R2 > 0.99) were found for all three regression models derived from the polygonal-derived dataset. However, all three models of the datasets derived from the alternative algorithm showed a poor correlation between retention time and (dependent variables of) mass/charge (m/z) for the alternative algorithm. To conclude, the use of linear regression models shows promising results, however, optimization of the alternative algorithm is needed prior application in the field of lipidomics. With the use of linear regression models and the alternative algorithm, we hope to optimize and automate lipid annotation for further studies.