Abstract: To solve the problem that the three-dimensional quantitative structure-activity relationship (3D-QSAR) model is not ideal when using the model to predict the biological activity of the new compounds, two new consensus models were established to improve the prediction ability of the model. A different weight to each submodule (named weighted consensus model, WCM) was added to one of the consensus models. In order to construct WCM, multiple linear regression (MLR) methods were used to calculate different weight coefficients for each submodule. Another consensus model was constructed from the average of the predicted values for each sub-model obtained in the literature (named average consensus model, ACM). Results show that the consensus model can improve the prediction ability when 0.5 < q² ≤ 0.8, but it can't improve the 3D-QSAR model's prediction ability when q² > 0.8. This result can help to improve the prediction of the model and the design of new high activity inhibitors.