jp_shionogi_logo


Virtual screening on the hundreds of millions of compounds has been realized.

2021

Shionogi & Co., Ltd. has been engaged in manufacturing and selling pharmaceuticals and clinical trial drugs since its founding. As a drug discovery-based pharmaceutical company, simulation using massive computing resources is indispensable. Historically they have used an on-premises environment, but due to the limited on-premises resources, their on-premises environment was not able to accommodate execution of their jobs. Therefore, they adopted the Amazon Web Service (AWS) to build an high performance computing (HPC) environment to flexibly secure the compute resources as needed. This has not only reduced calculation time by more than 10 times compared to their on-premises environment, but also helped them realize hundreds of millions of compound calculations.

AWS Case Study: Shionogi & Co., Ltd.
kr_quotemark

Shionogi has formulated a new mid-term management plan, 'Shionogi Transformation Strategy 2030 (STS2030).' Digital innovation is essential to transform ourselves from a traditional manufacturing company to a company that provides comprehensive healthcare services. We believe we can realize STS2030 by further promoting AWS utilization."

Hiroyuki Kobayashi
Shionogi & Co., Ltd. 
Vice President
Head of Digital Intelligence Department, DX Promotion Division

An HPC environment is essential in the drug discovery process

To “supply the best possible medicine to protect the health and wellbeing of the patients we serve" is the philosophy of Shionogi & Co., Ltd. The company, with infectious disease treatment drugs as its main product, has started a new coronavirus vaccine clinical trial. This is the first time a Japanese company has announced a clinical trial of a COVID-19 vaccine. The company is planning to prepare the vaccine production system for 30 million or more people by the end of 2021.

As it can take 9 -17 years and cost a tens to hundreds of billions of Japanese Yen to develop a new drug, and the success probability is said to be 1/25,000 (research by the JPMA [Japan Pharmaceuticals Manufacturers Association] ), streamlining the process is important. As it becomes more difficult to discover new chemical compounds year after year, large scale screening is required. In addition, to select a candidate for a new drug, massive data (such as genome data) analysis is a prerequisite and high speed computing resources are indispensable for that.

The company has implemented simulations to search candidate chemical compounds, however, with its traditional on-premises environment, if multiple researchers use the environment at the same time, job execution is a bottleneck. “Since the resources are limited on-premises, it takes many hours to calculate. It takes 6 months or longer to procure the resources as well. Since the personnel in charge of Research & Development has to manage the cluster, it has also affected the development work,” said Shin Kurosawa, Management Strategy Headquarters, Digital Intelligence Division, Future Creation Group and CCoE (Cloud Center of Excellence).

Realizing value in the cloud with AWS support

Shionogi has studied using the cloud to access massive HPC resources and adopted AWS after comparing multiple services.
“The deciding factor was that the AWS support system was very substantial. As it was the first time we planned to use AWS at full scale, we were initially anxious. AWS professional services and solution architect personnel very kindly taught us the necessary design patterns proactively. As they provided detailed documents based on their track record and practice that we could not obtain from other companies, we decided we could trust them with confidence.” (Shin Kurosawa)

Shin also said that he valued the availability of cutting-edge services and architecture for use. For example, AWS’ enhanced serverless and managed services helped alleviate the operation load, and AWS’ proven track record in the pharmaceutical industry, and abundant information in Japanese were all key differentiators.

The company prepared two types of AWS environments, one with emphasis on security and the other with emphasis on agility. The developers use the agile environment to test a newly released AWS service, and use the security environment to access internal data with high confidentiality. Furthermore, Shionogi adopted AWS Organizations that centrally manages AWS accounts, so that CCoE members can transparently manage 20 or more accounts securely.

Reducing time to insight for genome analysis

After building the AWS environments, they started developing use cases with the drug development team, and now multiple workloads have been conducted on AWS. HPC workloads include genome analysis, virtual screening, and three-dimensional protein structure analysis.

For genome analysis, they adopted AWS Batch to realize large-scale calculations by task to reconstruct the genome from a DNA fragment and to find the gene from the genome. They are utilizing AWS Batch according to the task, in the environment where the genome data volume can be tens of terabytes and the analysis pipeline is complicated. .
“Although we initially used Amazon EC2 for parallel computing, now we can easily scale by using AWS Batch. Although it used to take 3 hours to analyze one sample in the past, we can now analyze 100 samples simultaneously, thus we are able to complete the calculation several times faster than on-premises.” Keigo Masuda from the Innovative Medical Research Institute, Basic Research for Drug Development.

By utilizing EC2 Spot instances, they have saved on Amazon EC2 cost by 50% or more compared to On-Demand pricing. They further saved Amazon EC2 costs by about 30% by using the Oregon Region.

In virtual screening and three dimensional protein structure analysis, they use AWS ParallelCluster to build a cluster that automatically scales to search for a chemical compound that can be the drug candidate from hundreds of millions or more candidates. Shota Uehara from Advanced Medical Research Institute, Cheminformatics says, “Simulation was possible only within the limited calculation resources on-premises. Thanks to AWS ParallelCluster, we now can start a 1,000 CPU or larger cluster as needed, and the calculation that used to take one month or longer is now completed within 3 days. Because of this, we can try to search for the chemical compound at a scale.”

Furthermore, they have adopted NICE DCV, a remote access software, to confirm the results quickly and securely in the cloud. “NICE DCV provides a high-speed remote desktop environment, and operates the graphical analysis software that displays 3-dimensional molecular structure without delay.” (Shota Uehara)

Accelerating drug development with AI/ML

Utilizing AWS has enabled them to immediately deploy the massive resources required for drug development. They can efficiently drive their project by processing multiple calculations in parallel, which contributes to realizing the philosophy of the company, “to deliver medicines to people even a day sooner”. In addition, the company has realized indirect cost benefits, such as reduction of employee-hours required for infrastructure procurement and environment building.
“The benefit of AWS is that it can instantly respond to the needs of the researchers and reduces management cost only by combining multiple services. This is immeasurable at the Digital Intelligence Division where acceleration of drug development by digital technology is one of their goals. Speed is very important for our company to move forward as a drug development company, and I am convinced that AWS will continue to help us accelerate the process.” (Shin Kurosawa)

They plan to put a data lake in place on AWS for research and development and will expand their drug development activities by centrally integrating/managing open data and the data which is currently managed in a decentralized manner inside and outside the company. Furthermore, they are considering applying artificial intelligence/machine learning (AI/ML) managed services and utilizing them to detect an abnormal action by dynamic video analysis.

Because of this, they have high expectations on AWS, and Shin Kurosawa says “I would like AWS to continue supporting what a modern development method should be in the drug development field in the future.”

shionogi_casestudy_shugo
shionogi_kurosawa

Shin Kurosawa

shionogi_masuda

Keigo Masuda

shionogi_uehara

Shota Uehara


Customer Profile: Shionogi & Co., Ltd.

  • Date of establishment: June 5, 1919 (founded on March 17, 1878)
  • Capital: 21,279,742,717 JPY
  • Number of employees: 5,222 (as of March 31, 2020)
  • Business content: Pharmaceuticals, clinical test drugs and instrument research, development, manufacturing and sales, etc.

Benefits and future development with AWS

  • Improving development efficiency with AWS ParallelCluster and AWS Batch
  • Challenging the calculation at a scale not possible before with on-premises resources
  • Focusing on the original work of the researcher in charge of drug development pipeline by reducing management tasks
  • Optimization of infrastructure cost with pay-as-you-go billing and Spot Instances.
  • Contribution to the basic philosophy “To deliver medicines to people even a day sooner” of Shionogi Co., Ltd.

Key services being used

AWS ParallelCluster

AWS ParallelCluster is an open-source cluster management tool supported by AWS. This tool helps in the deployment and management of High Performance Computing (HPC) clusters in the AWS Cloud.

See here for details »

AWS Batch

Developers, scientists, and engineers can execute a few million batch computing jobs on AWS easily and effectively by using AWS Batch.

Click here for details »

NICE DCV

NICE DCV is a high performance remote display protocol that provides a secure method to deliver remote desktops and application streaming from the cloud and data center to any device under varying network conditions.

Click here for details »

AWS Organizations

AWS Organizations are useful to centrally manage and supervise the environment according to the increase in AWS resources and scaling.

Click here for details »