Classification of wood species in trade using metabolomic profiling by GC×GC-TOFMS

Abstract

Reports of illegal logging are increasing globally, driving a need for tools that can effectively identify wood products at the species level. This identification is crucial for regulatory purposes, certifying legal lumber, preventing environmental crimes, and protecting ecosystems and society. Current wood identification methods are primarily based on anatomical observation of wood tissues, chemical profiling using direct analysis in real-time time-of-flight mass spectrometry (DART-TOFMS), and DNA-based analyses. While these approaches have their advantages, they also present challenges, particularly when species-level identification is required for enforcement. As an alternative, metabolite profiling using separation techniques coupled with mass spectrometry shows potential as a robust species-level wood identification method. Here, we present a method for classifying wood at the species level through chemical profiling of the wood metabolome using comprehensive two-dimensional gas chromatography coupled to time-of-flight mass spectrometry (GC×GC-TOFMS) combined with chemometric analysis. In this study, different tissues, including sapwood, heartwood, and branches from seven wood species collected from genetically characterized trees representing three distinct genera (Quercus, Acer, Picea) were investigated. Principal component analysis (PCA) was used to visualize differences between species and tissue types, while partial least squares-discriminant analysis (PLS-DA) and feature selection were used to construct classification models for species-level wood identification. The classification models were built using data from wood cores, branches, or a mixture of wood cores and branch samples. Each classification model was tested with an external validation set, and the performance of the classification model was evaluated based on the prediction of the external validation data. Our results show that classification modelling using wood metabolomic data is promising, especially with the same tissue type, presenting accuracies of 100%, 100%, and 93.2% in the prediction of wood core samples at the species level for Quercus, Acer, and Picea, respectively.