A practical study of CITES wood species identification by untargeted DART/QTOF, GC/QTOF and LC/QTOF together with machine learning processes and statistical analysis

Abstract Illegal logging and trafficking of endangered timber species has attracted the world's major organized crime groups, with associated deforestation and serious social damage. The inability of traditional methodologies and DNA analysis to readily perform wood identification to the species level for monitoring has stimulated research on chemotyping techniques. In this study, simple wood extraction of endangered rosewoods (Dalbergia spp), amenable to use in the field, produced colorful hues that were suggestive of wood species. A more definitive study was conducted to develop wood species identification procedures using high-resolution quadrupole time-of-flight (QTOF) mass spectrometers interfaced with liquid chromatography (LC), gas chromatography (GC), and Direct Analysis in Real Time (DART). The time consuming process of extracting “identifying” mass spectral ions for species identification, contentious due to their ubiquitous nature, was supplanted by application of machine learning processes. The unbiased software mining of raw data from multiple analytical batches, followed by statistical Random Forest analysis, enabled discrimination between both anatomically and chemotypically similar Dalbergia species. Statistical Principal Component Analysis (PCA) scatterplots with 95% confidence ellipses were visually compelling in showing a differential clustering of Dalbergia from other commonly traded and lookalike wood species. The information rich raw data from GC or LC analyses offered a corroborative, legally defensible, and widely available confirmatory tool in the identification of timber species.