A Practical Approach for Validation of Compound Identification in Comprehensive Two-Dimensional Gas Chromatography–Mass Spectrometry

A simple curve-fitting approach for the retention index ( I ) calculation without reference compound injection was established. This was applied to validate compound identification data in comprehensive two-dimensional gas chromatography coupled with mass spectrometry (GC × GC–MS). Different models were investigated for calculating I in the first ( 1 D) and second ( 2 D) dimensional separations ( 1 I Cal and 2 I Cal, respectively) with the input data of compound names, elution temperature ( T e ), and experimental retention time ( t R, Exp ) of peaks, taking into account the wraparound effect. By comparing I Cal and the literature I ( I Lit ) of 495 compounds, a more suitable model was the exponential equation containing eight coefficients, which showed average R 2 of 0.9982 and 0.9957, mean absolute error (MAE) of 8.29 and 12.89, and root mean squared error (RMSE) of 11.63 and 16.48 for 1 I and 2 I comparisons, respectively. For compound validation, the initial assumption was that most of the peaks in a data set were identified with the correct compound names. The least-squares curve fitting was then performed to match 1 I Cal and 2 I Cal with 1 I Lit and 2 I Lit of these compounds. The compounds could be validated with correct identities when the fitting shows 1 I and 2 I differences (Δ 1 I = 1 I Lit – 1 I Cal and Δ 2 I = 2 I Lit – 2 I Cal, respectively) within ±60 units. Compounds showing |Δ 1 I | or |Δ 2 I | outside this range (observed with the coordinates out of the linear trendline of the I Cal vs I Lit plot) were suspected with incorrect identities. The developed approach was applied to validate 1741 peaks of 15 samples analyzed in this study. By using tentative identification based on “only MS” and “MS and 1 I ” identification criteria, 536 (out of 995) and 534 (out of 746) compounds, respectively, were suspected to have incorrect identities. Example reannotation of 61 peaks with more correct identities was also demonstrated. In addition, the approach was used to screen 1285 (out of 7445) suspected compounds in 85 GC × GC–MS results reported from the literature.