Impact of Data Scaling Strategies on Volatile Marker Selection in GC-MS-Based Classification of Single Malt, Blended, and American Bourbon Whiskies

Whisky volatiles exhibit an extensive dynamic range in peak areas, which causes highly abundant compounds to dominate multivariate models and mask minor flavor-active constituents. Thus, data scaling is critical because it directly modulates the contributions of different variables to the selection of markers and classification outcomes. We investigated the impact of unit variance (UV) and Pareto scaling on the volatile profiles of three types of whisky (single malt, blended, and American bourbon) using untargeted GC-MS. UV scaling maximized group separation but showed high sensitivity to minor components. Conversely, Pareto scaling provided a more chemically representative profile by preserving the inherent variance of major congeners, including ethyl esters, higher alcohols, and fatty acids. This methodology was validated by identifying key markers of fermentation, specifically ethyl caproate, ethyl caprylate, and ethyl caprate. Furthermore, standard-based identification confirmed that whisky lactones were significantly abundant in the American bourbon despite being undetected in untargeted profiling. Therefore, we conclude that reliable interpretation of complex whisky matrices requires aligning scaling with the analytical objective and combining untargeted profiling with standard-basd identification for trace sensory-relevant markers.