Evolving window factor analysis-multivariate curve resolution with automated library matching for enhanced peak deconvolution in gas chromatography-mass spectrometry fuel data.

Fuel chromatography is inherently limited by the high complexity of petroleum fuel compositions. In practice, almost no fuel components are fully resolved in gas chromatography. This is due to both insufficient peak capacity for the large number of individual components within time and chromatographic efficiency constraints, and insufficient resolving power of the stationary phase in the gas chromatography column relative to the many structurally similar isomers or homologs present in petrochemical fuels. Multidimensional approaches, longer columns and slower heating rates can offer some benefits but will not necessarily fully resolve co-eluting fuel compounds, especially within reasonable analysis times. The following work details how deconvolved mass spectral loadings, combined with library matching, provide a quality metric against which to automatically evaluate results obtained from an experimental evolving window factor analysis-multivariate curve resolution deconvolution algorithm applied to gas chromatography-mass spectrometry data. This algorithm was evaluated in the context of trace component detection in synthetic fuel data sets, dodecane and tetradecane detection in petrochemical fuels, and the detection of natural products unlikely to be present in petrochemical fuels. In the case of the trace component detection challenge, the experimental algorithm outperformed a control algorithm that utilized a singular value-based quality metric. Meanwhile, when detecting dodecane, tetradecane, and natural products in petrochemical fuels, the experimental algorithm allowed for higher-quality compound identification results than could be obtained without peak deconvolution, thus reliably improving fuel component resolution in an automated fashion.