Optimization of amphetamines multivariate detection by GC-FTIR spectra preprocessing

A chemometric method developed for the automatic detection of illicit amphetamines, which represent the most popular drugs in Europe, is presented. The detection system is based on Principal Component Analysis. The training database includes the GC-FTIR spectra of the main amphetamines, their precursors and derivatives. Although the system does not include any information on the biological activity of modeled compounds, the amphetamine analogues form two distinct clusters, according to their biological activity and toxic side effects (stimulants or hallucinogens). The clustering was optimized by preprocessing the spectra with a selective amplifier. This procedure increases significantly the efficiency of the detection system. The system correctly assigns the class identity of an unknown compound, even if its spectrum is not present in the database.