Estimation of the synthetic routes of seized methamphetamines using GC-MS and multivariate analysis.

ABSTRACT

One hundred and twenty six seized methamphetamine (MA) samples were analyzed using GC-MS. All the peaks that appeared in the chromatograms were investigated and 61 impurities including n-octacosane (internal standard) were identified. Among them, 37 impurities were already known or newly identified by comparing with commercial library entries and 18 impurities were detected for the first time. To estimate the synthetic routes of MA samples, route specific impurities had to be selected for each method. Two naphthalenes, 1,3-dimethyl-2-phenylnaphthalene and 1-benzyl-3-methylnaphthalene were selected as Nagai route specific impurities and three diasteromers, UK-19.62(58_165_178) I, UK-19.95(58_165_178) II, UK-20.49(58_165_178) III were also selected not only for their high frequency detection only in Nagai samples but also for the high principal component analysis (PCA) correlation values. For the Emde route, N,N-dimethyl-3,4-diphenylhexane-2,5-diamine and N-methyl-1-{4-[2-(methylamino)propyl]phenyl}-1-phenylpropan-2-amine were selected as route specific impurities, and N,N-di(ß-phenylisopropyl)amine I (DPIA I), N,N-di(ß-phenylisopropyl)amine II (DPIA II), N,N-di(ß-phenylisopropyl)methylamine I (DPIMA I) and N,N-di(ß-phenylisopropyl)methylamine II (DPIMA II) were selected for the Leuckart route. With these route specific impurities, synthetic routes could be identified for 78 of the 126 samples. The 61 impurities were registered in AMDIS target component library and the GC-MS data were deconvoluted. After AMDIS deconvolution, a matrix file was composed and then multivariate analyses were performed to estimate the synthetic route for unknown samples. The unsupervised methods, hierarchical clustering analysis (HCA) and PCA clustered the samples according to the closeness between samples. Two classification functions were obtained from discriminant analysis (DA) and the synthetic routes of the unknown samples were predicted using these two functions.