A multivariate analysis of the infrared spectra of drugs of abuse.

The infrared (IR) spectra of 15 drugs of abuse were analyzed for similarity by using techniques of numerical taxonomy. The study included six barbiturates (amobarbital, barbital, butabarbital, phentobarbital, phenobarbital, and secobarbital), four amphetamine-related compounds (amphetamine, ephedrine, methamphetamine, and phentermine), and five other drugs (cocaine, heroin, phencyclidine, phendimetrazine, and diazepam). Three character sets were based on increasing numbers (10, 24, and 36) of IR peaks. The cluster analysis, principal component analysis, and nonmetric multidimensional scaling elements of the program system NT-SYS were used to structure taxonomic distances between drugs. Best results were obtained from the 36-peak data set; ordination diagrams proved to be more visually informative than phenograms. Preliminary results from our analysis of this set of drugs indicate that an expanded multivariate approach to drug classification may be useful.