Cold plasma modulates the secondary structures of allergenic proteins in postharvest kiwifruit: An FTIR and machine learning analysis

Postharvest kiwifruit contains allergenic proteins that can trigger immunoglobulin E-mediated immunoreactivity, highlighting the need for strategies to modulate protein structure. This study aimed to integrate Fourier transform infrared spectroscopy (FTIR) and machine learning (ML) to characterize cold plasma (CP)-induced changes in kiwifruit protein secondary structure. FTIR with Amide I band deconvolution was used to quantify β-sheet, α-helix, β-turn, and random coil contents under CP conditions. CP treatment reduced the β-sheet content from approximately 50% in the control to 40%-43% under intensified treatment, accompanied by an increase in β-turn content. Two ensemble ML algorithms, random forest (RF) and gradient boosting decision tree (GBDT), were used to predict protein secondary structures from CP processing parameters. Overall, both models accurately predicted β-turn (R2 > 0.90) and α-helix (R2 ≈ 0.89), with slightly lower performance for β-sheet (R2 ≈ 0.81), whereas random coil prediction showed relatively lower accuracy. These results demonstrated the potential of ML to predict CP-induced protein structural changes, while SHAP analysis identified influential CP parameters, together supporting the selection of candidate processing conditions for targeted protein structural modulation. However, the effects of these structural changes on allergen immunoreactivity remain unclear. Future studies should integrate structural predictions with direct IgE-binding assays and other immunological assessments to clarify these relationships and guide immunoreactivity-oriented CP optimization in kiwifruit. Overall, this integrated FTIR-ML framework provided a methodological basis for targeted protein structural modulation during postharvest kiwifruit processing and offered a foundation for developing data-driven strategies for allergen management.