Comparison of PLSR, MCR-ALS and Kernel-PLSR for the quantification of allergenic fragrance compounds in complex cosmetic products based on nonlinear 2D GC-IMS data

• Co-elutions of analytes and matrix compounds prevent reliable quantification of potentially allergenic compounds in cosmetics by GC-IMS. • Competitive ionization situations of co-eluting analytes , and occurring multicomponent ion clusters lead to non-linear behavior. • Multivariate regression methods, such as MCR-ALS, PLSR and kernel PLSR, facilitate quantification drastically. • Reliable multivariate regression extends the field of application of GC-IMS to complex routine analysis.