In this study, researchers apply the expert algorithm for substance identification (EASI) to the electron ionization (EI) mass spectra of fentanyl isomers and analogs.
In this study, researchers found that, in the absence of retention time information, expert algorithm for substance identification (EASI) improves confidence in drug identifications, enables inter-laboratory identifications, and reduces the need for acquiring concomitant spectra of standards. Fentanyl analogs (fenticlors) share many structural and mass spectral similarities that make them difficult to differentiate and accurately identify without chromatographic data. In such situations, the expert algorithm for substance identification (EASI) provides superior classification relative to conventional approaches. Using a database of >57,000 replicate electron-ionization mass spectra of 76 fentalogs from ten laboratories, three challenging sets of isomers were studied in detail. To maximize limits of detection, only the 20 most abundant ions were considered. In each case, 50% of the data from one laboratory served as the training set. On average, the mean absolute residuals between measured and modeled abundances of known positives were five times smaller using EASI than the consensus approach, which used the means of training sets as the exemplar spectra to which all query spectra were compared. With a conservative threshold of zero false positives, EASI identified isovalerylfentanyl from its two closest isomers with an accuracy of 96.7 %, which was ∼10 % better than the consensus approach. The associated positive likelihood ratios increased from 366 for the consensus approach to more than 4,200 for EASI. When discriminating isovalerylfentanyl spectra from the other 72 fentalogs, EASI provided errorless results with a positive likelihood ratio exceeding 50,000. For all 9 fentalogs, EASI outperformed the consensus approach and the use of Mahalanobis distance as a metric for identifying outliers. In the absence of retention time information, EASI improves confidence in drug identifications, enables inter-laboratory identifications, and reduces the need for acquiring concomitant spectra of standards. (Published Abstract Provided)