This paper proposes a novel dynamic graph representation method called WiKG for histopathology whole slide image (WSI) analysis. The method constructs a dynamic graph representation of WSIs by modeling the interactions between head and tail embeddings of patches. It introduces a knowledge-aware attention mechanism that updates node features by learning the joint attention score of each neighbor and edge. The method then obtains a graph-level embedding through global pooling of the updated head, serving as an implicit representation for WSI classification. WiKG outperforms state-of-the-art WSI analysis methods on three TCGA benchmark datasets and in-house test sets. The method is effective in capturing the interactions between patches, leading to better WSI analysis performance. The method is evaluated on three public histopathological benchmark datasets and in-house test sets, demonstrating superior performance in cancer typing and staging tasks. The method also shows strong generalization ability on in-house frozen section lung cancer WSIs. The method is compared with other state-of-the-art WSI analysis algorithms, and the results show that WiKG achieves superior performance in terms of AUC and F1-score. The method is also effective in capturing the directional contributions between entities, allowing more valuable entities to propagate more practical information. The method is efficient in establishing topological structures and has a fast convergence rate. The method is also effective in capturing the interactions between patches, leading to better WSI analysis performance. The method is evaluated on three public histopathological benchmark datasets and in-house test sets, demonstrating superior performance in cancer typing and staging tasks. The method is also effective in capturing the directional contributions between entities, allowing more valuable entities to propagate more practical information. The method is efficient in establishing topological structures and has a fast convergence rate.This paper proposes a novel dynamic graph representation method called WiKG for histopathology whole slide image (WSI) analysis. The method constructs a dynamic graph representation of WSIs by modeling the interactions between head and tail embeddings of patches. It introduces a knowledge-aware attention mechanism that updates node features by learning the joint attention score of each neighbor and edge. The method then obtains a graph-level embedding through global pooling of the updated head, serving as an implicit representation for WSI classification. WiKG outperforms state-of-the-art WSI analysis methods on three TCGA benchmark datasets and in-house test sets. The method is effective in capturing the interactions between patches, leading to better WSI analysis performance. The method is evaluated on three public histopathological benchmark datasets and in-house test sets, demonstrating superior performance in cancer typing and staging tasks. The method also shows strong generalization ability on in-house frozen section lung cancer WSIs. The method is compared with other state-of-the-art WSI analysis algorithms, and the results show that WiKG achieves superior performance in terms of AUC and F1-score. The method is also effective in capturing the directional contributions between entities, allowing more valuable entities to propagate more practical information. The method is efficient in establishing topological structures and has a fast convergence rate. The method is also effective in capturing the interactions between patches, leading to better WSI analysis performance. The method is evaluated on three public histopathological benchmark datasets and in-house test sets, demonstrating superior performance in cancer typing and staging tasks. The method is also effective in capturing the directional contributions between entities, allowing more valuable entities to propagate more practical information. The method is efficient in establishing topological structures and has a fast convergence rate.