The paper introduces MINES, a novel framework for inductive relation reasoning in knowledge graphs (KGs). MINES addresses the limitations of existing GraIL-based models by introducing a Message Intercommunication mechanism on Neighbor-Enhanced Subgraphs. The key contributions of MINES include: 1. **Message Intercommunication Mechanism**: This mechanism captures omitted hidden mutual information by introducing bi-directed information interactions between connected entities through an undirected/bi-directed GCN layer inserted between undirected RGCN layers. This enhances the expressive ability of the model. 2. **Neighbor-Enhanced Subgraph Extraction**: The model extends the neighborhood area beyond the enclosing subgraph to include more isolated neighbors, which are proven to be beneficial for inductive relation reasoning. 3. **Extensive Experiments**: MINES demonstrates superior performance, effectiveness, and transferability on various inductive datasets, outperforming existing models in terms of AUC-PR and Hit@10 metrics. 4. **Ablation Study**: The study shows that the message intercommunication mechanism and the neighbor-enhanced subgraph extraction are both effective and necessary for improving the model's performance. 5. **Transfer Analysis**: The proposed approach is scalable and generalizable to other GraIL-based models, as demonstrated through experiments on TACT and CoMPILE. 6. **Conclusion**: MINES provides a more powerful and expressive framework for inductive relation reasoning in KGs, offering improved discriminative and expressive capabilities.The paper introduces MINES, a novel framework for inductive relation reasoning in knowledge graphs (KGs). MINES addresses the limitations of existing GraIL-based models by introducing a Message Intercommunication mechanism on Neighbor-Enhanced Subgraphs. The key contributions of MINES include: 1. **Message Intercommunication Mechanism**: This mechanism captures omitted hidden mutual information by introducing bi-directed information interactions between connected entities through an undirected/bi-directed GCN layer inserted between undirected RGCN layers. This enhances the expressive ability of the model. 2. **Neighbor-Enhanced Subgraph Extraction**: The model extends the neighborhood area beyond the enclosing subgraph to include more isolated neighbors, which are proven to be beneficial for inductive relation reasoning. 3. **Extensive Experiments**: MINES demonstrates superior performance, effectiveness, and transferability on various inductive datasets, outperforming existing models in terms of AUC-PR and Hit@10 metrics. 4. **Ablation Study**: The study shows that the message intercommunication mechanism and the neighbor-enhanced subgraph extraction are both effective and necessary for improving the model's performance. 5. **Transfer Analysis**: The proposed approach is scalable and generalizable to other GraIL-based models, as demonstrated through experiments on TACT and CoMPILE. 6. **Conclusion**: MINES provides a more powerful and expressive framework for inductive relation reasoning in KGs, offering improved discriminative and expressive capabilities.