Multi-fields model for predicting target-ligand interaction
Predicting target-ligand interactions is a critical task of chemogenomics and plays a key role in virtual drug discovery.Moreover,it is important to lake insights into the molecular recognition mechanisms between chemical substructures of ligands and binding sites of targets.In this work,we suppose the interaction between a ligand and a target is the result of the comprehensive effect of multiple fields between the ligand and the binding site of the target,and propose a multi-field interaction model(MFIM)to predict the target-ligand interaction.The evaluation result on the same data set shows that MFIM outperforms other two representative methods.The derived fragment interaction network is robust to the parameter fluctuation and the connections in the network are sparse.Moreover,the edge weights of the network might reflect the fragment interaction intensity and most of the significant edges are chemical inlerpretable.
Protein representation Ligand representation Field interaction Target-ligand interaction Target-ligand network analysis
Caihua Wang Juan Liu Fei Luo Qian-Nan Hu
School of Computer,Wuhan University,Wuhan 420072,PR China Tianjin Institute of Industrial Biotechnology(TIB)of Chinese Academy of Sciences(CAS),Tianjin 300308
国内会议
大连
英文
604-611
2016-09-24（万方平台首次上网日期，不代表论文的发表时间）