|
内容記述 |
An mD-VcMD simulation is a generalized ensemble method developed by us recently to study ligand?receptor docking. The simulation starts from completely dissociated conformations of the ligand and the receptor, samples various complex conformations in an explicit solvent, and generate a free-energy landscape covering both bound and unbound conformations. Here, we propose a novel version of mD-VcMD, maned “cartesian-space-division-mD-VcMD” (or CSD-mD-VcMD), which can predict the ligand?receptor complex structure without knowledge on the ligand?receptor binding site, although the previous mD-VcMD version (referred to as “DSD-mD-VcMD” in this study for convenience) requires the knowledge. We applied the current method to the same system studied by DSD-mD-VcMD: A ligand (ribocil A or ribocil B) binding to an RNA molecule (the aptamer domain of the FMN riboswitch), with assuming that the system’s binding sites are unknown. As a result, the free-energy landscape of ribocil B was funnel-like, and the ligand was assigned to the deep binding pocket of the aptamer domain in the lowest free-energy basin. In contrast, the landscape for ribocil A was ragged-like. We report that the two methods provides similar results. As known, RNA?ligand complex-structure prediction is a difficult task because RNA and ligand are in general highly flexible. Therefore, one cannot definitely answer if an experimentally proposed complex may involve a large structural error, or there is a possibility that the complex is a structure temporally emerging in large structural fluctuations. CSD-mD-VcMD may propose an important suggestion to this difficulty. In fact, the current study showed that a few complex clusters exist in the deep ligand-binding site, and one of the clusters was the native-like and the other are stabilized by inter-molecular stackings and/or inter0-molecular hydrogen bonds between the ligand and receptor. |
