Apples and derivate products are rich in polyphenols, the consumption of those specialized metabolites has potential health benefits. Through apple juice processing (crushing pressing), the oxidation naturally occurs caused by the contact of the plastidial polyphenol-oxidase (PPO) and its vacuolar phenolic substrates. Coupled to oxygen presence, the PPO catalyzes the oxidation of apple native polyphenols to numerous oxidation products (OPs) with specific structures and functional properties. Organoleptic (color, bitterness, astringency, turbidity) and nutritional apple juice qualities might therefore be modified by oxidation processes. To better understand its mechanism and consequences on juices and ciders, it is necessary to know the molecules formed by oxidation. In contrast to previous works dealing with targeted analysis of OPs on the basis of oxidative reactivity of polyphenols [1,2], the present study aims to set up a standardized non-targeted approach for detecting Ops in juices of different apple varieties depending on the amount of oxygen consumed . Juices from five cider apple varieties were prepared under anoxic conditions. They then underwent a controlled oxidation by cumulative additions of oxygen reached thanks to an experimental prototype. Thus, five oxidation levels (null to maximum) were analyzed for each experiment by HPLC-UV-MS. Samples were injected before and after depolymerization by phloroglucinolysis to characterize both native and oxidized procyanidins. Raw mass spectrometry data were processed using Galaxy Workflow 4 Metabolomics (W4M) to obtain the full ions list. An original workflow was developed using a four steps data filtering to retain OPs related ions. A linear regression model (polynomial 2nd order) was applied to each ion, variety by variety to fit the intensity of each ion to the cumulative injected oxygen. The model enabled (i) the collection of all ions that responded to oxidation, (ii) the removal of weak oxygen dependent ions, (iii) the selection of ions that increase with the supply of oxygen (OPs). Then the last step (iv) used a hierarchical agglomerative clustering to annotate ions (retention time, molecular ion, isotopes, fragments, adducts, stacking) into “molecular signals”. This approach enabled the identification of 32 OPs previously identified by targeted approaches and 22 potential OPs that had not been detected before. The presented non-targeted approach associated to quantitative sorting is strengthened through the detection of already known OPs and even provided supplemental information about OPs. Ultimately, the approach will contribute to the deep understanding of oxidations processes and their effects on the juice quality and composition. [1] Castillo-Fraire C.M., Pottier S., Bondon A., Salas E., Bernillon S., Guyot S. & Poupard P., Food Chemistry, 372, 131117, https://doi.org/10.1016/j.foodchem.2021.131117, 2022. [2] Mouls L. & Fulcrand H., Journal of Mass Spectrometry, 47, 1450-1457, https://doi.org/10.1002/jms.3098, 2012.