In this study, 44 AH samples were subjected to a quantitative analysis using SWATH, and 557 proteins were identified in each sample according to a well-established library of 634 proteins. Proteomic profiles in AH of IU and VKH groups were significantly different from normal controls, whereas relatively minor inter-group variation was observed. Differential analysis revealed a shared pattern of extracellular matrix disruption in IU and VKH groups and, more importantly, downregulation of retinal cellular proteins highlighting the value of AH as a medium for ocular disease research. Enrichment analysis showed a typical inflammatory state in the protein composition of AH in IU and VKH groups compared to normal controls, where innate immunity played an important role, as indicated by the activation of the complement cascade and the overexpression of innate immune cell markers. Subsequently, an efficient and robust machine learning algorithm, XGBoost, was used to develop a model for classifying IU, VKH, and normal controls. The top three ranked proteins in the importance list, TF, ENPP2, and FUCA1, were found to represent a promising biomarker panel.