This paper proposes a novel method for predicting drug-target interactions (DTIs) based on signed heterogeneous graph neural networks (SHGNNs). The method models DTIs on signed heterogeneous networks, taking into account both the positive and negative effects of interactions between drugs and targets. The approach involves categorizing interaction patterns of DTIs and extracting interactions within drug pairs and target protein pairs. The proposed SHGNN-DTI framework not only adapts to signed bipartite networks but also incorporates auxiliary information from drug-drug interactions (DDIs) and protein-protein interactions (PPIs). The framework addresses the message passing and aggregation problem on signed DTI networks and considers different training modes on the whole networks consisting of DTIs, DDIs, and PPIs. Experiments conducted on two datasets extracted from DrugBank and related databases show that the proposed method achieves excellent performance in terms of metric indicators. The feasibility of the method is further verified by a case study with two drugs on breast cancer. The results demonstrate that the SHGNN-DTI framework is effective in predicting signed DTIs and can provide valuable insights into the comprehensive mechanisms of drug combinations. The method is able to handle the complexities brought by both DDIs and PPIs information, and it is more robust in terms of embedding dimensions. The study also shows that the cooperative mode is better than the independent mode, and that SHGNNs with sharing weights are more robust in terms of embedding dimensions. The results indicate that the SHGNN-DTI framework is an effective way to predict signed DTIs and that DDIs and PPIs can provide benefits for promoting performance of SHGNNs.This paper proposes a novel method for predicting drug-target interactions (DTIs) based on signed heterogeneous graph neural networks (SHGNNs). The method models DTIs on signed heterogeneous networks, taking into account both the positive and negative effects of interactions between drugs and targets. The approach involves categorizing interaction patterns of DTIs and extracting interactions within drug pairs and target protein pairs. The proposed SHGNN-DTI framework not only adapts to signed bipartite networks but also incorporates auxiliary information from drug-drug interactions (DDIs) and protein-protein interactions (PPIs). The framework addresses the message passing and aggregation problem on signed DTI networks and considers different training modes on the whole networks consisting of DTIs, DDIs, and PPIs. Experiments conducted on two datasets extracted from DrugBank and related databases show that the proposed method achieves excellent performance in terms of metric indicators. The feasibility of the method is further verified by a case study with two drugs on breast cancer. The results demonstrate that the SHGNN-DTI framework is effective in predicting signed DTIs and can provide valuable insights into the comprehensive mechanisms of drug combinations. The method is able to handle the complexities brought by both DDIs and PPIs information, and it is more robust in terms of embedding dimensions. The study also shows that the cooperative mode is better than the independent mode, and that SHGNNs with sharing weights are more robust in terms of embedding dimensions. The results indicate that the SHGNN-DTI framework is an effective way to predict signed DTIs and that DDIs and PPIs can provide benefits for promoting performance of SHGNNs.