<p>Price-aware recommendation aims to predict items a user is likely to purchase by considering both item prices and user historical interactions. Current methods typically discretize prices into uniform or logistic-based levels, representing price with a single value. However, real-world price distributions are highly complex, making it difficult to model them accurately with simple distributions or categorization. Additionally, existing approaches focus mainly on extracting user price preferences based on item prices, overlooking the fact that user behavior is influenced by both price and interest preferences, which have a complex interdependence. To address the above challenges, we propose a novel method called Price-Aware Graph-Based Recommendation Using Latent Price Vectors (PaGReL). First, we introduce a latent price vector constructor based on a Gaussian Mixture Model (GMM) to effectively model and represent complex price distributions, generating a latent price vector for each item. Next, we leverage a Graph Convolutional Network (GCN) to extract embeddings that capture semantic information from both user-item interaction graphs and item attribute graphs. Compared with graph models such as LightGCN that ignore explicit price factors, PaGReL achieves more refined economic dimension user modeling through a dual-channel graph encoder. Finally, we model the intricate relationship between price and interest preferences using user and category embeddings, dynamically adjusting the importance of each preference for specific item categories. Extensive experiments on two real-world datasets validate the effectiveness of the proposed PaGReL method. In summary, the contribution of this article is to construct a latent price vector through Gaussian mixture model (GMM) and to balance price sensitivity and user interests using PAD.</p>