Ratio-Consistency-Driven Discrimination of Structurally Similar VOCs in PTR-MS

Plant volatile organic compounds (VOCs) provide important chemical information related to plant physiological states and plant−environment interactions. Proton transfer reaction mass spectrometry (PTR-MS) enables real-time VOC monitoring with high sensitivity, but overlapping protonated and fragment-ion signals limit the discrimination of structural isomers and mass-similar VOCs. This study proposes a ratio-consistency framework for VOC discrimination under fixed E/N conditions. Centered log-ratio (CLR) transformation is used to represent relative fragment-ion abundance relationships, while a ratio-consistency constraint guides the model to exploit reproducible fragmentation-related patterns and reduce its reliance on absolute intensity variations. For eight structurally similar VOCs, the proposed framework improves closed-set classification accuracy from approximately 55% using intensity-based features to 89.2%. An open-set recognition strategy combining maximum Softmax confidence and ratio-consistency deviation is further introduced, achieving an overall recognition accuracy of 81.1% across known and unknown samples. The framework is subsequently applied to VOC emissions from mechanically wounded tomato plants, where PTR-MS analysis identifies β-pinene as a major VOC released during the early stage of wounding, with independent FTIR measurements providing consistent spectroscopic support. These results demonstrate that relative fragment-ion abundance patterns provide complementary discriminative information for PTR-MS analysis and offer a practical strategy for improving the discrimination and open-set recognition of structurally similar VOCs without additional instrumental modulation.