Real-world Knowledge Graphs (KGs) are inherently incomplete, necessitating effective link prediction to infer missing facts and enhance their utility in knowledge-driven applications. While embedding-based methods falter in inductive scenarios with unseen entities, Graph Neural Network (GNN)-based link predictors sacrifice interpretability despite their expressive power. Path-based alternatives address these limitations through transparent reasoning patterns, yet conventional implementations relying on Closed Paths (CPs) between entity pairs face severe path scarcity in sparse KGs. To overcome this bottleneck, we propose Open Path from Pruned Graph enhanced Link Predictor (OPPGLP), a novel framework that integrates GNN-derived relational patterns with path-based reasoning. By converting GNN’s latent features into semantically meaningful paths, our approach achieves both explainable and accurate link prediction. Comprehensive evaluations across three public datasets demonstrate state-of-the-art performance of our OPPGLP in 5 out of 6 transductive and inductive settings.