Toray Industries, Inc.

Background

Digital Annealer expected to solve complex optimization problems

The limit to increases in computer performance, as expressed in Moore’s Law^*1 , is fast approaching. It is anticipated that quantum computing will provide the solution to this, becoming the next generation of computing architecture. However, few actual solutions are currently available. Fujitsu’s Digital Annealer is one such solution, designed for solving combinatorial optimization problems by adopting a general purpose solution called the ‘simulated annealing’ method. It brings advantages of quantum computer into the digital circuit. Fujitsu is already conducting joint research with a number of companies to discover groundbreaking ways of using the Digital Annealer in business, and Toray Industries (Toray) is one of partners.

The first theme of the joint research between Toray and Fujitsu was the optimization of molecular structures. They verified predictions of the most stable structures of proteins by using Digital Annealer. According to Dr. Ryuji Tanimura, Head of Computer-Aided Drug Design, Pharmaceutical Research Laboratories at Toray, “Proteins, which include enzymes, antibodies and receptors, form complex structures and experimentation to analyze them is extremely labor-intensive.”

To determine the structure of a protein through experimentation, crystals with stable molecular structures must be created and analyzed with X-rays or other techniques. However, crystallizing and stabilizing proteins is a difficult process. Growing protein crystals is a time-consuming process. It can take anywhere from a few months to half a year, or even more in some cases and it is not uncommon to discover, even after investing all this time that the process has failed. Consequently, to overcome such lengthy intervals, researchers are looking to computation as a means of predicting the optimal combinations of different, novel molecular arrangements, what is known technically as the side chain^*2 conformation^*3 of proteins.

Dr. Tanimura explains, “Combination algorithms capable of predicting optimal structures have been available before. The performance of conventional general-purpose computers has allowed us to use these to make predictions for small proteins, but as the combinations of the side chain conformation increase, it was hard to reach answers in practical time.”

Toray looked to the Digital Annealer as a means of potentially overcoming this problem.

• *1
  Moore’s Law is a prediction that the number of elements in an integrated circuit would double every 18 months in Semiconductor industry
• *2
  A spot where branches from main chain of proteins
• *3
  The spatial arrangement of the atoms affording distinction between stereoisomers which can be interconverted by rotations about formally single bonds

Outcome and next step

A shift to “in silico” is changing the rules of the market competition

This co-creation project between Toray and Fujitsu provided confidence in the ability of the Digital Annealer to design proteins, so the development of a design program for predicting desired protein structures is being considered as the next step. Meanwhile, Toray and Fujitsu are currently planning their next joint research project to conduct trials in the area of new materials, which accounts for a large share of Toray’s sales. The two companies have also begun to investigate the formulation of new algorithms to use the Digital Annealer for simulating the development of polymers and other materials.

Digital Annealer used in this project had 1,024 bits connectivity, and will evolve into a second generation with 8,192 bits, which will expand the scope of its application. It is also expected to be able to solve more complex problems.

Dr. Tanimura explains his vision, “In the past, drug candidates were evaluated randomly, with verification through animal testing at the end. There can be up to a million possibilities. By narrowing numbers of evaluating methods down through the computation as much as possible, we hope to reduce the number of possibilities down to 1,000 or even 100.” This represents a shift from in vivo (animal experiments) through to in vitro (test tube experiments) and now to in silico (computer experiments).

This shift changes the principles of research and development of new medicine competition. “If we can improve the precision of in silico, it will become a contest of intelligence rather than scale and size. That is the world that I wish to see,” concludes Dr. Tanimura enthusiastically. Fujitsu and Toray have launched a co-creation initiative that will open frontiers and bring us to the next level.