Implementation of an Advanced Workflow to Enhance the Confidence in Compound Identification for Non-Targeted GC-HR-MS Applications

Chemical characterization still remains a major hurdle for analytical chemists when dealing with non-targeted screening analyses, despite significant improvements in chromatographic separation techniques and mass spectrometric instrumentation over the last decade. This work focuses upon the monitoring of volatile and semi-volatile compounds using gas chromatography coupled to high resolution mass spectrometry (GC-HR-MS) in electron ionization (EI) mode, using both headspace and liquid injection modes. A total of 559 reference compounds, including odd n-alkanes (n=5 to n=19) as chemical markers, were analyzed and experimental linear retention index (LRI) values were determined. A personal compound database library containing EI accurate mass spectra and experimental LRI values was created to enable rapid screening and accurate identification for compounds present in aerosol samples. The reference compounds were used to create computational Quantitative Structure-Property Relationship models using two independent approaches: RapidMiner combined with Dragon software and ACD/ChromGenius software [1, 2]. These models were then used to predict LRI values for several thousand compounds reported to be present in tobacco and tobacco aerosol, plus a range of specific flavor compounds [3]. Using these two prediction models, the mean predicted LRI values for all compounds falling within 500 to 1,900 were calculated and used in conjunction with EI nominal mass spectra (from NIST & Wiley libraries) to enhance the confidence in structural elucidation during complex matrix analysis.