It is important to understand how proteins consistently fold into their native-state structures and the relevance of structure to their function. In this work, we study the unfolding pathway of SH3 from the perspective of network. Our results show that the average clustering coefficient is less sensitive to the structural change and the average shortest path lengths can examine the larger structural changes. The betweenness values of the hydrophobic cores locate at the local maximum in the network of native state and that of the folding nucleus locate at the maximum in the network of transition state. It is easy for the betweenness of the transition state network to distinguish the folding nucleus from other residues and identify the hydrophobic core regions from the betweenness of the native state network. The hydrophobic core and folding nucleus all become derogated from native state to denature state. The hydrophobic collapse model and nucleation condensation model become reconciled in this finding.