To establish a compound-specific isotope profiling method by liquid chromatography-high-resolution mass spectrometry (LC-HRMS) for identify caffeine and lovastatin derived from natural extraction, chemical synthesis and bioengineered fermentation in food matrices.

Waters XBridge BEH C₁₈ column (250 mm×4.6 mm, 5 μm) coupled with a mild elution gradient was used to moderately broaden the chromatographic elution window of target compounds. A quadrupole-orbitrap mass spectrometer operated at a resolution of 140000 full width at half maximum (FWHM) was used to continuously acquire the complete isotopic peaks of target molecules at no fewer than 20 non-equidistant time points within each chromatographic peak. During data processing, a theoretical correction model was applied to subtract heteroatom isotope contributions. Grubbs’outlier test and the pruned exact linear time (PELT) change-point detection algorithm were then used to quantify intra-peak isotopic heterogeneity and stepwise shifts. Full-elution deconvolution was further used to construct a two-dimensional isotope landscape model of the food matrix.

Caffeine from natural tea sources, chemical synthesis and bioengineered fermentation, as well as lovastatin from natural red-yeast fermentation and pharmaceutical-grade sources, showed distinct baseline isotopic fingerprints. In simulated adulteration systems,when exogenous synthetic substances accounted for 10% of the total target compound, the PELT algorithm accurately detected isotope-ratio change points at the P<0.01 confidence level; When the proportion was greater than or equal to 20%, stepwise transition points were reliably localized. The adulteration proportions back-calculated using the isotope mass-balance model showed relative deviations within ±15%. Over the concentration range of 0.1–10.0 μg/mL, the 5 consecutive days of testing gave an inter-day relative standard deviation of 0.8% for δ¹³C, with a maximum systematic bias of +0.8‰, demonstrating high precision and strong resistance to matrix interference.

The multidimensional isotope-profile time-series method established in this study enables qualitative differentiation between natural and artificially synthesize forms of the same molecule by exploiting chromatographic isotope fractionation without requiring complete chromatographic separation. This approach provides a feasible technical solution for resolving source-identification challenges involving substances with overlapping natural occurrence and artificial addition, and offers a methodological reference for improving food safety regulatory strategies.