DUAL PARALLEL DETECTION RAW DATA FUSION: QUANTITATIVE FOOD VOLATILOMICS ON LARGE SAMPLE SETS

Volatilomics, a rapidly expanding discipline, aims to characterize volatile metabolites in various matrices. Advances in two-dimensional gas chromatography (GC×GC) coupled with parallel flame ionization detection (FID) and mass spectrometry (MS) offer combined capabilities of compound identification (by MS) and precise quantification (by FID). Parallel detector signals are generally processed separately, this study integrates MS and FID chromatograms to enhance pattern recognition during template matching, enabling comprehensive quantitative volatilomics. Feature matching utilizes MS spectral similarity, reducing mismatches and facilitating accurate FID response extraction for quantification. Applied to hazelnuts, a premium confectionery ingredient, GC×GC-FID/MS analyses identified volatiles distinguishing cultivars, geographic origins, post-harvest treatments, contamination, oxidative stability, and sensory quality. This methodology aligns with the Sensomics-Based Expert System (SEBES), an AI-driven platform predicting food aromas without human olfaction. The dataset, encompassing raw hazelnut samples over four harvest years, required normalization of MS responses and correction with internal standards due to variability. Temporal chromatographic misalignments can cause 2D peak pattern inconsistencies. Data fusion, guided by MS spectral similarity, reduces false negatives by about 80% for 441 detectable features compared to FID alone and minimizes false positives, enhancing specificity and selectivity. Additionally, data fusion halves processing time and facilitates metadata transfer. After pattern recognition, FID signals are extracted for quantification based on calibration curves or predicted FID response factors. This combined detection method for quantitative volatilomics reliably tracks aroma changes over crop production and shelf-life, enabling robust marker discovery for industrial quality assessment.