12Motivation: The interactions of proteins with DNA, RNA, peptide, and carbohydrate play key roles in 13 various biological processes. The studies of uncharacterized protein-molecules interactions could be 14 aided by accurate predictions of residues that bind with partner molecules. However, the existing 15 methods for predicting binding residues on proteins remain of relatively low accuracies due to the 16 limited number of complex structures in databases. As different types of molecules partially share 17 chemical mechanisms, the predictions for each molecular type should benefit from the binding 18 information with other molecules types. 19Results: In this study, we employed a multiple task deep learning strategy to develop a new 20 sequence-based method for simultaneously predicting binding residues/sites with multiple important 21 molecule types named MTDsite. By combining four training sets for DNA, RNA, peptide, and 22 carbohydrate-binding proteins, our method yielded accurate and robust predictions with AUC values of 23 0.852, 0836, 0.758, and 0.776 on their respective independent test sets, which are 0.52 to 6.6% better 24 than other state-of-the-art methods. More importantly, this study provides a new strategy to improve 25 predictions by combining multiple similar tasks. 26 Availability: http://biomed.nscc-gz.cn/server/MTDsite/ 27Contact: