Boosting pattern recognition specificity in GC×GC volatilomics through FID/MS signal fusion

Volatilomics is an emerging discipline aimed at characterizing volatile metabolites in various samples, with applications ranging from the food industry to biological research, serving as a powerful tool in food quality and authenticity assessments. Comprehensive two-dimensional gas chromatography (GC×GC) coupled with flame ionization detection (FID) and mass spectrometry (MS) combines structural information (MS) with robust quantification (FID). Traditionally, data from the two detectors are analyzed separately. The challenge tackled in this study has been the fusion of MS and FID chromatograms to improve the reliability of pattern recognition during template matching while enabling quantitative volatilomics on a large set of detectable features. The process of features matching across chromatograms is guided by the spectral similarity provided by MS information, thereby reducing mismatches while, at the same time, extracting the FID responses for accurate quantification. This approach was applied to hazelnuts, a premium ingredient in the food industry, to investigate their aromatic profiles influenced by cultivar, geographic origin, post-harvest treatments, microbial contamination, oxidative degradation. This strategy aligns with the Sensomics-Based Expert System (SEBES), which act as an AI-Smelling predicting food aromas without human olfaction. Hazelnut samples collected over four production years exhibited variability in MS response and chromatographic misalignments, leading to inconsistencies in 2D peak patterns. The application of data fusion, driven by MS spectral similarity, minimizes feature mismatches, reducing false negatives across the 441 detectable compounds in raw hazelnuts when compared to FID alone and lowers false positives, thus improving the method’s specificity and selectivity. Data fusion also halves processing time and facilitates metadata transfer. After pattern recognition step, FID signals are extracted for quantitation based on calibration or predicted FID response factors. Quantitative volatilomics by parallel detector signal fusion tracks aroma blueprint changes over crops and along the shelf-life, while enabling robust markers discovery for industrial quality assessment.