Laboratory of Informatics of Molecular Functions(AIST) Recruiting students for the academic year 2018

Associate Professor Takatsugu HIROKAWA
E-mail: t-hirokawa{at}
Lab HP


【Key Words】Protein 3D-structure modeling, Protein-ligand docking, Pharmacoinformatics, Molecular Dynamics

 Structure-based drug design (SBDD)―in which new pharmaceuticals are designed based on the three-dimensional structure and known interactions of a target receptor protein ―has attracted increased attention due to the development of structural genomics. Yet conventional computer-based SBDD requires precise 3D coordinates for the target protein structure, and compound docking simulations require highprecision structural search and interaction energy calculations. However, there are almost no crystalline structures in pharmaceutical target protein families such as G-protein coupled receptors (GPCRs), so there are high expectations for computerized methods of molecular modeling. While other common target proteins such as the tyrosine kinase family have relatively many crystalline structures,structural deformations near the binding compound call for optimization and evaluation of the structure’s suitability for SBDD.
 Against this background,we conduct research with an aim at developing methods for molecular modeling of target proteins and docking simulations, as well as virtual screening methods.

Molecular modeling of target proteins

 We are investigating methods of molecular modeling specific to the structure of the target protein, focusing on methods for comparative modeling and molecular dynamics simulations for the GPCR and tyrosine kinase families. For example GPCR is a target protein with seven transmembrane helices, and we are attempting predictions from sequence analysis of the amino acid residues required for stable existence between helices, reflecting the results in structure predictions. Figure 1 shows an example of molecular modeling for a histamine receptor.

Fig .1:Molecular modeling of a human H2his tamine receptor. Projection is in the horizontal direction from the membrane (green).

Compound modeling

 There are many previously reported methods for simulating docking between proteins and compounds, and many software packages that can perform such simulations. However the precision of docking simulations depends on the protein or compound,and other problems remain,such as searching for candidate binding compound structures and needed improvements to compound evaluation functions. We are developing a method called CoLBA for evaluating compound binding by simulating active site docking for representative compounds such as inhibitors after target protein structures are formed. CoLBA is advantageous in that considerations of candidate structures obtained through docking simulation occur not only based on interaction energy,but also by using results from multiple compounds to mutually compare molecular interaction profiles with the target protein,thereby arriving at a consensus determination of the binding state.This allows flexible and intuitive screening that does not depend on interaction energy alone(Fig.2).

Fig . 2:A binding model for three types of inhibitor for a his tamine HI receptor, selected by CoLBA.

Virtual screening

 Virtual screening is performed based on molecular modeling of a target protein and a binding model between the target protein and a compound, as obtained from a compound docking simulation. The hit ratio is simulated for a group of compounds known to be active toward the target protein, and a group of non-active compound randomly selected from a library. The results of this evaluation are fed back into molecular modeling of the target protein and in the docking simulation process, helping to optimize the target protein􌖨compound model.
 In future joint research we hope to evaluate physiological activities by selecting compounds with models created from libraries of millions of compounds.


The University of Tokyo
Graduate School of Frontier Sciences, The University of Tokyo

Page Top