Cloud-based Supercomputer Leads to Evolution in Drug Discovery

Development of a new drug requires at least 10 years and enormous expenses, but only one in 30,000 candidates are put into practical use. To improve these conditions, the University of Tokyo’s Research Center for Advanced Science and Technology (RCAST) and Fujitsu has joined forces to embark on in silico drug discovery. By creating a high-precision simulation environment that makes full use of the cloud and supercomputers, RCAST aims to improve prediction accuracy to eliminate unnecessary experiments and increase the efficiency and success rate of new drug development.

The probability a new drug is put into practical use is only 1/30,000

Medicine is indispensable to our lives. In particular, there are increasingly higher expectations for new drugs with significant benefits in the fields of preventive medicine, geriatric medicine, intractable disease treatment, and advanced medicine.

However, development of a new drug generally requires 10 to 15 years of development time and enormous research and development expenditures. The process starts by finding out which substance (protein) affects the illness being studied and identifying which protein causes the illness. Then, a compound or antibody is developed to control protein activity in order to suppress the onset or progression of the illness. Drug candidate compounds are created through this process of drug discovery. Between 2006 and 2010, over 670,000 candidate compounds were created in Japan, but only 22 were approved as medicines (Japan Pharmaceutical Manufacturers Association investigation). Thus, the probability of a new drug being put into practical use is extremely low— only about 1 in 30,000.

One reason for this is that in conventional drug discovery, synthetic chemists often modify and improve a known compound in order to create drug candidate compounds, so it is difficult for them to create new drugs which have an efficacy unavailable from known compounds. Another reason is that even if a drug candidate compound is created, often problems emerge later during a clinical trial, which reveals the ineffectiveness of the compound as a drug or gives rise to concerns about its side effects. Clinical trials take an immense amount of time and are expensive; in addition, if they fail, researchers must repeat the process of designing, remaking, and verifying a compound all over again. Creating candidate compounds for groundbreaking new drugs that cannot be found with traditional techniques and predicting its effectiveness and side effects accurately before moving into clinical trials have been critical challenges in new drug development.

To solve these challenges, Fujitsu has been collaborating with the University of Tokyo's RCAST to tackle in silico drug discovery, which makes full use of supercomputers. Fujitsu is conducting research and development of new drugs to treat cancer and other diseases by designing compounds through computer simulation.

Creating basic compounds for new drugs through simulation

In research on new drugs that may enable treatment according to individual patients' conditions, it is necessary to understand the complex in vivo activities of the proteins thought to cause diseases in more detail. This requires processing a huge volume of calculations, far more than has been done previously. Could not a large-scale, high-performance simulation environment be built on-campus? To respond to this request of RCAST, Fujitsu provided a customized version of our cloud service for analytical simulations, the FUJITSU Technical Computing Solution TC Cloud (TC Cloud).

By using the cloud, RCAST has secured powerful computing capabilities. In addition, the cloud environment also allows RCAST to conduct joint research with pharmaceutical companies and other organizations. Moreover, due to power consumption restrictions on campus, a supercomputer that uses a large amount of electricity would be limited in terms of scalability, but our cloud service permits RCAST to build a large-scale computing environment not subject to electricity restrictions.

Creating highly novel candidate compounds through simulation; ex-ante prediction of effectiveness eliminates unnecessary experiments

The University of Tokyo's RCAST expects to accelerate new drug development by establishing a new simulation environment. Exhaustive searching using a high-performance computer has enabled RCAST to virtually create highly novel candidate compounds digitally, which would not have been possible from refining known compounds. Moreover, high-precision simulation of protein activities and candidate compounds in the human body now allow RCAST to predict the effectiveness of candidate compounds in advance in order to eliminate unnecessary experiments.

RCAST started full-scale use of TC Cloud for new drug development on January 1, 2014. TC Cloud will contribute to the steady, revolutionary development of highly effective anticancer drugs with few adverse effects as well as other new drugs, which were previously difficult to put into practical use with traditional experimental drug discovery techniques.