Theme 2 Simulation Applicable to Drug Design (Representative: The University of Tokyo, Professor Hideaki Fujitani) searches for candidate compounds that strongly bind to target proteins of diseases using the absolute binding free energy computation method based on the originally developed molecular dynamics (MD) method. Furthermore, they analyze interaction between proteins aiming at renovation of the drug design process by the computational science. This time, we interviewed two researchers who are involved with the studies on new chemical compounds which suppress functions of target proteins of diseases and design of antibody drugs utilizing the K computer.

Encounter with the biosimulation research

●First, I'd like to hear about how you two started working on the present research.

Yamashita :  I have worked on the computational science since I was attracted to describing physical phenomena in mathematics in my high school days and worked on the computational science. I was interested in quantum mechanics and studied on proton. Having found a lot of phenomena greatly affected by the proton in the field, I was engaged in the biological field such as the molecular biology in the US and studied on proteins. That is my background. I was interested in drug design using protein simulation. I thought the simulation would be applied more widely than I expected, so I decided to challenge this theme. The research directly connected to medical service is related to priority issues in the society, so that is worth doing. I am really happy if I can contribute to the society. I joined the University of Tokyo in January, 2011. Since then I have studied on biosimulation and the molecular dynamics calculation that targets proteins related to drugs.

Shinoda :  I was very attracted to the phenomenon called “phase transition” in physics in my college days. So, I worked in an experiment laboratory first, and was involved with studies mainly with X-ray diffraction experiments of alloy using synchrotron radiation. I had a chance to observe phenomena that could rarely be seen in the experiments and I was very excited with it. But, at the same time I hoped to dig into model researches and theory to find how and why such phenomena occur. I took a master course and a doctoral course of different graduate schools. I studied on phase transition of diluted ferromagnet in the master's course and on quantum liquid based on an integral equation theory in the doctoral course. I was involved with a research with the MD simulation for the first time when I was a post-doctoral researcher of National Institute of Advanced Industrial Science and Technology. After that, I had worked on various themes such as material development or quantum chemical calculation as a post-doctoral researcher of enterprises. Meanwhile, I had been in half-hearted attitude all the time. I could take time to hypothesize and carefully by myself. I could not dig into my research, which made me frustrated. I was employed by the University of Tokyo in December, 2010, a month before Dr. Yamashita was. With MD simulations of biomolecules using the supercomputer and examination of their structures and dynamics, I researched to develop drugs. Because the researches including examination of their structures and dynamics to support drug design have its clear purposes and direct relation to the output. Such researches really interested me and were worthwhile doing. The professors of the wet team working with me are eager to develop drugs with the results of computational science. The results of simulation research are directly applied for drug design, which gave me a great pressure on me; however, I will make an effort to do it.


Promote research that contributes to new drug design

●What researches are you two working on for Theme 2?

Yamashita :  What I first dealt with Theme 2 was a research on molecular target drugs that inhibit kinase.
Many drugs have an effect by suppressing functions of the proteins that cause the disease. In order to discover a drug, it is necessary to find out chemical compounds which strongly interact with the target protein in vivo. Therefore, we attempt to evaluate absolute binding free energy of target proteins and chemical compounds by simulations with a supercomputer so as to find chemical compounds that can be used as drugs. The absolute binding free energy is the physical quantity that indicates how much molecules can recognize target proteins. We predict the activity of new candidate compounds with high precision by the method of the absolute binding free energy computation, "MP-CAFEE", developed by Prof. Fujitani and his colleagues. We design and search for chemical compounds with high inhibitory activity. The MD program, the engine of "MP-CAFEE", is "GROMACS". It has been developed as an open source program. We tuned this up for the K computer in cooperation with developers of Stockholm University and others. The computation improved approximately twice as efficiently as the primary version. This speeding up allows the absolute binding free energy computation to be performed more efficiently. In this way, a technical base to find out promising candidate compounds has been established. Large-scale data began to be accumulated since last year. It enables us to write papers that examine possibility of new drug design. Now our important mission is to analyze the computation results and provide outcomes as science.
 I have been doing a basic research, and my greatest motivation is in there while I am attracted to mysterious aspects of basic phenomena and things not clarified when doing the applied research for drug design. While application progresses steadily toward a practical use, there actually are a lot of thing remaining insufficient in basic parts. Not all are completed as perfect methods. As long as we belong to Academia, we need to continue the research further. In order to achieve it, we perform actual computation widely and obtain results, continuing researches for such aspects.


Shinoda :  I have studied molecular dynamics for antibody drug development against cancer. The use of antibodies that bind specifically to antigens has achieved considerable success in cancer therapy in recent years. Antibodies are applied to the methods such as pre-targeting method, which involve separating the targeting antibody from the subsequent delivery of therapeutic agent that binds to the tumor-localized antibody. Now we try to improve the affinity of current treated antibodies and investigate basic mechanism: a specific way of binding antigens - antibodies by simulations. This is because the antibody treated by us is very unique. There is a proline in the region where antigens are recognized, and its conformation changes from cis to trans by binding to antigen (cis-trans isomerization). This transition requires great amount of energy and it is said that it assumes a switching role for various biological processes. It takes too much time to see the mechanism of this transition by the MD simulation directly; however, if we can obtain a clue to understand the mechanism, it can be applied to the design of the part for recognizing antigens, we believe.
 In order to clarify the binding processes and structures, we performed nearly 1,000 MD simulations using the K computer, and obtained various trajectories of antigen binding to antibody. In other words, we intend to estimate the mechanism observing various cases by performing the extensive simulations. It is biologically essential to clarify such an underlying mechanism, which leads to drug design in the future.

Building up research vision for the next generation is also a mission

●Going into the final year, how do you wish to advance your research based on the results obtained so far?


Yamashita :  As I have just mentioned, a method for high speed computation being established so we are challenging a large-scale computation of over 200 chemical compounds for one theme. In the future, it will be important to brush up the method to elucidate "Why" through more detailed analyses. I'm sure drugs such as the kinase inhibitor will be designed and explored efficiently. Further, the experiment is important for applying simulation results to drug design, I believe. No matter how good the simulation result is, what simulation results can show is only imitation of the reality and nothing but prediction. In order to get close to the real phenomena, it is necessary to conduct proper experiments and verifications at various phases, so as to verify consistency between the simulation and the reality. I believe that repeating such processes will lead to realization of highly accurate simulations and also be helpful for improving the speed and efficiency of drug design process.
 About the relation between experiments and simulations, I suggest that young researchers should understand that simulations are not performed to agree with experimental results. Models themselves are based on physics and built up purely from the theory. My remarks "experiments are important" doesn't mean that simulations results are to be adjusted to make them consistent with experimental results conveniently but means that we find some parts of the reality that the model can’t reproduce. Experiments are, so to speak, something like "checking the answers" and are important to find out why the simulation results do not agree with experimental results, which is very important and the researchers must be done for in the simulation research.
 Thinking of those things, I’m going to reflect the major theme that I have been in charge of in order to research outcome, adding the simulations to respond to new questions while brushing up methods and techniques. The simulation results are given to the researchers for the next experiments but the beneficial of the results is in my hands. So, if we can make use of them for the improvement of the future analysis techniques, or acquire hints to understand interesting characteristics of proteins from them, the research will be even more meaningful and interesting. Not only using the methods that have already been established, but also making them improving with our knowledge is fantastic. In some cases, the absolute binding free energy is not the only way and there may be other techniques by which better results are obtained in combination with the absolute binding free energy. Algorithms can also be improved based on a new viewpoint. We have mainly focused on speed-up but I believe there must be other approaches that we can attempt. For instance, the variation of effective candidate compounds can be widened by changing the approaches for the activity prediction. They may be worthless in the immediate future but I believe it is also our important mission to propose the ideas and theories that can be utilized in the next generation or next challenge.

Shinoda :  Last year, we performed approximately 1,000 MD simulations in an antigen- antibody system containing water with different initial velocity, and analyzed binding process and structure, water dynamics, and so on. We continue the analysis now, and our job for this year is to summarize the analysis results, I suppose. Dr. Yamashita has just mentioned that "GROMACS" has come to run in the K computer. I actually have received its benefit, which made my job much easier. However, as the MD computation was tough work, we paid great attention to it. In particular, the K computer in the early stage was not always capable of giving 100% of the expected data. It caused some problems such as half-way system freezing and data crashed by errors. At last conducting nearly 1,200 simulations, we could acquire approximately 1,000 date out of them. Further, data being obtained successfully, we needed to check if they were complete being. Anyway, as it was my very first time to perform such a massive computation at the beginning, we had no idea which part to check in the data, so as to draw out data sets bringing outcome. Finally, we managed to finish checking, and then at least approximately 1,000 of them were all right. The analysis still going on, we've found that the time of binding, duration and frequency of the binding vary in each case and that some do not even bind. Thus we have observed various behaviors. I presume that such a massive simulation had never been performed for a binding process of antigens and antibodies so far. In this sense, we carefully examine the dynamics of the binding process from these results.

Wish to make use of the research results for medical service

●Look forward five and ten years later, what researches do you wish to work on in the future?

Yamashita :  Thinking of the cycle of a research, it usually takes about five years obtain answer of the research. So, ten years are not a long period. As I said at the beginning, being interested in physical mechanisms, I am not going to continue only biology researches.However, because molecular behavior in proteins and cells have been clarified one after another and the biosimulation is making great progress, I find this field very interesting. In this sense, in the future we are going to focus not only on one single protein but perform simulations for a protein complex and focus on proteins playing special roles. I continue exploring interesting physical phenomena based on the results of such researches. Of course, as long as we treat biomolecules, we cannot ignore actual medical service so I try to perform a simulation research to support it. A research to clarify the cause of a disease is a good example. After all, the clues seem to be hidden in various diseases. I think slight difference in proteins causes unexpected serious symptoms in some cases. I am sure that it will be more important in the future to reveal what small errors in design or assembly should cause at molecular or atomic levels. I want to contribute to that.


Shinoda :  At present I have only treated only one kind of antibody. I want to is a kind of be a workman, or a specialist who knows all about simulations of antibodies in five years. Experts of the quantum chemistry would find errors intuitively looking at the molecular structure, and said to me "there's something wrong with it. I feel something strange in this result". Surprisingly I asked him why he knew that, and he said "you'd see it putting yourself in the molecule's position".
 I wish I would say, "Make it little moved, and we'll get an antibody to bind better" only looking at visualized images of molecular simulation. Further, through the research of antibodies, I do my best to clarify the mechanism of important phenomena, which contributes to the medical service. Of course I work on researches on the computational science with the computer simulation technique, however at the same time I have a feeling to face the biology or other approaches such as statistical dynamics, which is still a vague idea.