− How did you feel when SCLS started?

Yanagida: Actually, when I was appointed as Director out of the blue, I felt a little embarrassed, asking myself why I was appointed as Director in spite of my career as a specialist of single-molecule measurement, and the fact that I had never been professional in computers. I remember that, in my first speech, I said: "I doubt if computational science will ever be so beneficial to life science" (laugh).

Kidera: Yes, I remember that (laugh).

Yanagida: I think I made such a speech because I had an impression that something had been achieved in the field of life science without any deep knowledge, and I felt no need to know complex processes. However, as a matter of common sense, life scientists were empirically thinking that it was time for a change, and obliged to ask themselves what they should do to overcome such a situation. Meanwhile, measurement technologies such as genetic analyses did not stop advancing, and the complexity of data was increasing beyond our understanding. Thus, as the momentum for recognizing the importance of computing increased, I gradually started to gain an understanding. Besides, not only I but others too started to accept "the need to understand complicated things without simplifying them." Traditionally, a researcher simplified a complex phenomenon in his head and did an experiment in line with the hypothesis to find a solution. Every scientist pursues such a method. However, life science is not that simple. No mechanism in that field can be expressed by one simple mathematical formula. More complex analytical work would be required. As a result, we need a computer. Thus, I started to think "Computational science is indispensable for research activities in life science." But when we started, we recognized the sheer lack of computer power. In spite of that, when the members started to make efforts to use the K computer, we recognized the potential of the K computer for giving birth to a science which is different in kind, and the Director changed his mind and got earnest about promoting our efforts. All this happened thanks to the efforts made by those in computational science.

Kidera: The first speech of Dr. Yanagida was so impressive and striking. If I remember correctly, I said: "Please don't say something which would discourage the researchers" after his speech (laugh). His statement was so striking that I couldn't help giving him a warning. Why? Because his statement was right on the mark. In a word, at that time, computational science did not offer any results which would truly contribute to life science. SCLS was preceded by the "Next-Generation Integrated Simulation of Living Matter (ISLiM)," in which we basically conducted software development. Although its achievement forms the basis of this project, we were engaged only in software development but not allowed to use the K computer, so we did not get to attain any achievements which would convince life scientists. Besides, no clear or specific attention was directed toward the kinds of targets, calculation methods or desired achievements. So, as to the achievement of ISLiM, we could only say "We succeeded in developing general-purpose software. It's amazing, isn't it?" As we knew that, we did not want Dr. Yanagida to make a statement which would reiterate that. However, once we started to use the K computer, we marveled at its potential, and found our specific targets and what we should see, while the research system to achieve them was prepared. As researchers knew how they should use the K computer to do computations, they became experienced to some extent and confident, and finally were no longer afraid of what Dr. Yanagida says. That was when the project was halfway finished. At last, the image of life science based on the supercomputer was emerging, and achievements were being made little by little. After that, research activities progressed more rapidly by gaining more momentum. We have had five years so far.

Yanagida: In that sense, I succeeded in my first speech (laugh).

Kidera: Let alone succeeding… Anyway, the strongest impact was your first speech and various organizational changes at the intermediate stage. That is, you stressed the importance of thinking how to collaborate with life scientists who are outside the field of computational science. I believe that the project got rolling smoothly only because of such efforts. Before then, though researchers were not withdrawn into the shell of computational science, they succumbed to a vague and indefinable sense of understanding by creating an excellent piece of software without definitive products. But now, it is different.

Yanagida: No, actually, I said more gently: "Computers are interesting, but they will be more interesting if you apply them to understand the mechanism of life together with life scientists" (laugh). In this way, a new science was born, which was different in kind.

Eguchi: As Dr. Yanagida said at the beginning of the project, unlike other fields, only little-known researchers use computers in the field of life science. In a word, this is still a minor field quantitatively as well as qualitatively. The researchers were not sure about what they should do in this situation. That has been true. But, I wonder if it is still true even now. What should we do under such circumstances? In my opinion, the most important thing is to connect those researchers involved in research activities using the K computer with the research sites of universities and private companies.

I always worked with the hope that those universities and industries approach and see the researchers who use the K computer and discover its usability, potential for drug development or feasibility in medical settings, and they consider to use it by seeing it working. I started like that. Specifically, one successful project used "MP-CAFEE" developed by Dr. Fujitani. At first I consulted Dr. Kidera, and he answered: "I think that 'MP-CAFEE' would be friendly to everyone. The others are too professional and unfriendly to laypeople. "Consequently, a group was formed for industries based on "MP-CAFEE." At first, it was joined only by two pharmaceutical companies. But now, it has been joined by more than 20 companies. Maybe, such a "connection" between university researchers is very important, and I believe it is essential for outreach activities. Industries and research institutions including universities are taking advantage of computational science more frequently on a global scale. Needless to say, as far as the field of life science is concerned, such trend has not been spread widely in Japan. To promote the development of such a connection, we have to foster the next generation of those researchers involved in the current project and the one subsequent to that. In that sense, education is very important. Considerable efforts are required. But fundamentally, such a role should be fulfilled by universities or high schools, I guess.

Yanagida: In life science, I understand that university education is insufficient as computational science is still immature. But in other areas, a computer is already mandatory as fundamental technology, that is, scientific technology. Still, there aren't enough software developers, whereas there are a lot of users.

Eguchi: I guess that it isn't easy to spotlight it as an area of science.

Yanagida: It is like the development of a technology. No matter how great the efforts one makes to develop a fundamental technology, it's always those who applied it and realized an interesting or attractive achievement who stand in the spotlight.

Eguchi: That's also true in cases where a piece of existing software is ported to the K computer. To exploit the potential of the K computer to the full, a life scientist has to write a new program. But he/she would say "That's not my line. There should be someone else who takes care of that."

Yanagida: You mean researchers want to do life science research activities, not computer research activities, right? After all, university reform is necessary. Although this is not a matter we should go into, while new fields are being explored one after another and students want to do research activities with state-of-the-art technologies, there is no Department of Computational Life Science in universities. I guess efforts should be made to offer education that covers writing of software programs for computers, and their development into an exciting science.

Kidera: Life science and medical science have long histories. Now we have computational science and information science. However, there is no field connecting them. Why not? Because every university organization fundamentally consists of faculties, and thus, interdisciplinary areas between such faculties do not grow easily. To cope with this situation, for example, you establish a research institute dedicated to the research of interdisciplinary areas at university to cultivate talents. But, needless to say, you cannot attract students. With no human resources and no progress in studies, you are eventually obliged to engage in such studies on your own, isolated. I think this is the typical situation of universities in Japan. I often hear that interdisciplinary areas are not growing in every field including computational life science.

− I would like to ask what kinds of achievements have been made during the past five years with regard to the four Themes of SCLS.

Kidera: In each of the four Themes, achievements have been made, which are slightly different from each other. In the ISLiM, the need for software development was first advocated. However, what was required actually resulted in achievements which would be significant to life science.

The most significant theme at that time was the interconnection of the hierarchical layers of life. Life consists of spatially small to large components, that is, molecules to cells, organs and the entire body, and temporally short to long phenomena on a tremendously wide scale. We were required to make their interconnections understandable in detail. I truly felt it would be impossible, as the K computer was still unavailable. But the only way was to keep trying. After the start of SCLS, calculations on the K computer level were made possible for the first time, and the way to connect the hierarchy of life began to open up, although it was only in its beginning. For example, for the molecular-level research of Theme 1, efforts were made to handle the biological phenomena of cell scales containing a lot of proteins and biomolecules, and a simulation, which may be the first cell model accurately reproducing the behaviors of proteins and so forth in cells, is about to be developed.
Theme 3 is a challenge to the hierarchical connection on a larger scale. In this Theme, the heart simulator called "UT-Heart" was developed by miraculously establishing a connection from the molecular scale to the organ, heart. This is truly a marvelous achievement. Needless to say, if molecules are handled at the same level as Theme 1, even the K computer's capacity cannot cope with them. However, by adopting the same spirit, a model representing the behaviors of upper cardiac muscle cells was successfully developed, resulting in a very accurate and sophisticated whole heart model. Furthermore, Theme 3 tackled a more challenging theme, that is, to connect the brain, nerves, muscles and bones, and built an entire body model, which is now about to be put into practice. For Theme 3, I believe that not only the hierarchical connection was realized based on a simulator consisting of diverse algorithms on small to large scales, but also an integrated hierarchical simulation was realized. By making full use of the K computer, the answer to ISLiM's issue, that is, hierarchical connection, is sufficient. As a result, we are finally getting ready to respond to various questions concerning life science.
As Dr. Yanagida told us at the beginning, in the field of traditional life science, connections including the hierarchical one were forcibly built in many ways. The genotype and phenotype are good examples. In Theme 4, large-scale computations were carried out by using the K computer for analyzing a huge quantity of data to thoroughly research them and find out their relationships, focusing on the fact that the phenotype is more complicated. Such large-scale life data analysis is now about to enable a connection between the genotype and the phenotype in a real sense, and clarification of their complicated system. Not only has the data analysis become large-scaled. The phenotype is information on diseases from clinical sites, while the genotype is information on patients of those diseases. Efforts are being made to build a system by gaining the trust of medical doctors on site in that diseases might be understood by connecting the genotype and the phenotype, and to enable their connection in the true sense. This is truly the most marvelous thing. It is highly valuable that an infrastructure was built on which information obtained by the next-generation sequencer in clinical practice is being utilized as truly useful information.
Theme 2 was built around Dr. Hideaki Fujitani, who was invited to respond to direct social demand: "What is the supercomputer used for? Achieve results!" I was surprised by him at first, who responded to the question "Can the K computer develop a drug?" by saying "It's worth trying." I was secretly wondering if he was really sure. However, contrary to my expectation, he ultimately achieved three outcomes which led to the preclinical stage including an antibody. Of course, I believe he gained much support from various parties, but he successfully demonstrated the fact that such a variety of results can be obtained by use of the K computer for a limited time. This is a great achievement. His attitude in pursuit of maximum efficiency also greatly stimulated this strategic field. The researchers involved in the project including us had much to learn from him.

Yanagida: Above all, I would like to stress the fact that the computer power of the K computer made the achievement of such outcomes possible. Some may think that, even by using a low-performance computer, such outcomes could have been achieved with more time. That is not true. It is evident that every Theme has tackled issues which could not have been solved without the computer power of the K computer. As I said earlier, the project demonstrated that computational science was changing the quality of life science. I think that is the important point.

Kidera: For the hierarchical connection, for example, we actually want an infinite computer capacity. However, the capacity is limited. Researchers are always thinking what to do within such limitations. The computation method depends on the upper limit. However, with the advent of the K computer, computational flexibility, as well as the range of potentials to be pursued, increased. I think this is truly significant. In future, we will have much more opportunities. The K computer is, so to speak, a prototype which will bring such opportunities.

Eguchi: I have a feeling that these four research groups are fundamentally changing the concept of life science. In recent years, we have witnessed astounding progress in measurement technologies, and we can now obtain a huge amount of data at one time. As a consequence, there is growing demand for a method to describe a phenomenon in an integral manner based on a large amount of data, as in the case of Theme 4, instead of conventional life science in which attempts are made to understand a phenomenon by focusing on a certain part of the data obtained. The K computer responds to such demand. I am confident that unprecedented research styles which were not realized before will emerge one after another.

Yanagida: Quite a few life scientists still believe that computers are a kind of tool which support and demonstrate their experimental results and tentative theories. However, computational science is not just to support life science. I hope they understand that computational science is pioneering life science. It is in a position to lead data-driven life science. Computational life science does not advance unless you delve into data analysis technologies as well as methods for studying analytical results rather than mere computations. That is, life scientists are responsible for exploring such fields. To this end, I hope that those with potential for understanding the most advanced software programs tackle life science, and make full use of the K computer to explore new data-driven life science fields. Computational science is already a part of life science.

− In the past five years, you have had hard times, haven't you?

Eguchi: For the first two or three years, researchers repeated the process of trial and error for using the K computer. They had hard times. Before the advent of the K computer, ISLiM started software development, and based on that, researchers at SCLS proceeded with their respective development activities. However, they had struggled　in drawing the maximum performance out of the K computer because they had to engage in high-speed and large-scale computations beyond comparison with conventional ones.

Kidera: There were cases where we were obliged to retry a computation from the beginning which had been done by spending more than six months. Even though we were asked to do the computation while checking it, that was practically impossible, as we had to proceed with our research activities, tackling and organizing a huge amount of data in the belief that the computation results are always correct. It would be a "mission impossible" to find technical errors while doing such tasks. At any rate, it is very hard to use the software program on the K computer, get buried in a sea of data, reproduce the experimental data, and discover something new. We have to achieve an outcome by processing a large amount of data and doing time-consuming tasks at the same time. Therefore, it is no exaggeration to say that we repeated the process of trial and error by the middle of the project. But we are confident that we eventually succeeded in developing what is called authentic research prototypes for all the four Themes by doing a lot of computations and spending much time. I believe that we created a model for future researchers even if they have a new system or a different theme.

− How do you plan to build on the achievements obtained in the past five years?

Kidera: I think that research activities and their outcome using the K computer will sprout forth buds when computer resources increase further in future. We successfully built a solid model for conducting research activities, that is, how to accept a large amount of information, how to handle it, and what outcome should be achieved. I hope that more researchers will participate in this field.

Yanagida: What was achieved by the K computer is not necessarily sufficient. If a computer equipped with a computer power 100 times greater than the K computer is developed, we will want to do the computations to connect hierarchies further, and to solve more general problems. For example, we may want to understand how the heart works as part of the entire physical system, or how it is coordinated with mental status. I was a university physiology professor, and the most important area for medicine is physiology. For example, we want to know the physiological conditions of the entire body, such as how the mental status controls the immune response. I am confident that computational life science will pave the way for the understanding of this aspect, too. This will revolutionize the fields of life science and medicine.

Eguchi: I think it is measurement technologies and computers that are going to change the field of life science. However, even if a computer increases its capacity 100 times, high-precision calculation would not be possible without the development of measurement technologies. Measurement technologies also have to evolve in response to the development of computational science. For instance, in order to simulate an entire cell or an entire heart in the true sense, the measurement technologies responding to the concept of "entire cell" or "entire heart" must be developed accordingly. Otherwise, it would simply end up as a deskbound discussion. Measurement technologies and computers will need to progress in tandem by responding to each other.

Yanagida: You are right. Modern measurement technologies do not measure a complex thing as is. Under this hypothesis, a target phenomenon is focused on, simplified, and then measured. So, around 99% of data is discarded, as only the data satisfying the hypothesis is left. A measurer will provide 100 times more data than at present provided that any measured parameter of the entire body can be evaluated by the computer. Also, with such a system, a measurer will be able to provide required data accordingly and enter it into the computer for subsequent computations. It will also become possible to survey the reaction after the introduction of a drug in detail by focusing on the target through computation and measurement. Measurement methods will change, and the quality of life science will also change, which will drive a paradigm shift.

Kidera: In my opinion, data assimilation, which enhances simulation accuracy by taking experimental data into a model, will be recognized increasingly as a key concept in the field of life science. For this purpose, at the front of life science, researchers are now required to use computers. That is to say, there's no need to open up the field of computational life science. It can be integrated into the field of life science. It would be rather ideal.

Eguchi: You mean that computers and software programs are becoming like conventional test tubes and reagents. If the time comes when we can see computers in a life science laboratory as ordinary items, we will no longer need to use the words "computational life science," right?

Kidera: I think it was the first project to draw words like "We cannot do anything without a supercomputer any more" from some researchers at the site. I hope that the number of such researchers will increase.

Yanagida: I am sure that the time will come when life scientists make full use of computers like test tubes.

Open Up Special Talk : Approach to the four Themes during the past five years and its achievements

  Theme 1  Simulations of biomolecules in cellular environments
  Theme 2  Simulation applicable to drug design
  Theme 3  Hierarchical integrated simulation for predictive medicine
  Theme 4  Large-scale analysis of life data