Threpp (Multimeric Threading based Protein-protein Interaction Predictor) is a computational algorithm for protein-protein interaction (PPI) prediction. Starting from a pair of protein sequences, it does two things: (1), it will judge whether the two proteins interact with each other by calculating the likelihood through a naive Bayes classifier model which combines the Threpp threading score and available high-throughput experimental (HTE) data. (2), it creates the quaternary stuctural models of the PPIs by reassembling the monomeric threading templates with the identified PPI frameworks. Large-scale benchmark tests showed that Threpp can significantly improve the precision and recall of both HTE and multimeric threading, and therefore reduce the false positive rate for the current PPI modeling approaches. The performance of the current Threpp server is optimal for predicting PPIs in E. coli, for which the integrated HTE datasets are constructed. In case that HTE data is not available, Threpp will only use the dimeric threading score to assess the PPI likelihood.