Biological networks show various types of biological relationships which respond to various biomédical contexts, such as diseases and drugs. Therefore, the construction of a context-specific biological network is very important to those seeking to discover the underlying biological principles with respect to biomédical contexts. In recent studies, attempts to construct a context-specific network by determining the biological relationships from very large public biological databases and from the literature have been emerged. Therefore, finding biological relationships of which the contexts are related to specific contexts of interest is an important research issue. However, there are limitations in previous studies in that they only found a limited range of contexts and only considered pathological features from ontology. In this research, two strategies are employed to overcome these limitations. First, continuous similarity for all context term pairs is measured and biological relationships which have contexts with higher levels of similarity than an established similarity threshold are included in a context-specific network. Secondly, molecular biological features stemming from the similarity between context-associated protein-protein interaction networks is employed and integrated with the pathological similarity from ontology. The performance of the proposed method was evaluated by F-score comparisons and by the ROC curves, with the result showing higher overall performances than those of previous methods. The context-specific biological network in this research will promote additional research on the underlying biological principles of biomédical contexts.