Agile development has recently been garnering attention as a method that involves quickly ascertaining needs and progressing with a project in an agile and flexible manner. Chugai Pharmaceutical is home to the Tech Workshop, a team responsible for agile in-house application development. In this article we hear from members of the team that developed an app to automate the process for confirming the crystallization of drug substances. This was the first project that the Tech Workshop was involved in. This discussion sees drug discovery researchers and members of the Tech Workshop speaking about the effects of agile development and what they learned in the process.

*Reproduced from Chugai Pharmaceutical’s official Note page (https://note.chugai-pharm.co.jp/) Article details and employee positions are current as of September 2024

In the production of pharmaceuticals, what is the crystallization process?

In the production of pharmaceuticals, what is the crystallization process?

 

Ukegawa: The crystallization of drug substances is a key process in ensuring the stable mass production of high-quality pharmaceuticals. Drug substances are the active ingredients in pharmaceuticals. In the pharmaceutical production process, crystallizing of drug substances in a solution removes most of impurities to produce high-purity active ingredients.

While there are various crystallization methods, researchers are constantly exploring the optimal conditions based on different experiments.

 

The difficulty of crystallization depends on the molecular weight of the compound that is to be crystallized. Generally, small-molecule compounds (with a molecular weight of not more than 500) are easier to crystallize, and around 30 experiments are usually enough to discover the crystal. On the other hand, mid-size molecule compounds (between 500 and 2,000) are much harder to crystallize. Crystallizations may not be achieved even after over 1,000 experiments while changing various conditions, and the number of combinations of conditions to examine is increasing rapidly.

What was the impetus for developing an app to confirm crystallization?

Tanida: Although we already boast world-class drug discovery technologies for antibody drugs and small-molecule drugs, we are currently focusing on mid-size molecule drugs as our third therapeutic modality. We set up this project to reduce the vast number of experiments required for mid-size molecule drug discovery* and to streamline the process for confirming crystallization images.

For us to deliver pharmaceuticals to patients across the world as quickly as possible, it was essential that we made use of an approach based on machine learning.

How does the app work?

Goto: This app makes use of something called deep learning to confirm crystallization. We fed data of more than 500 each of crystallized and non-crystallized images that had been visually confirmed by researchers, and using this image data the model can automatically identify the features necessary for the confirmation of crystallization.

Goto: These are actual images from the app. Crystallization has been confirmed in the images with the red strips. The app also displays the certainty of each judgement by the model, supporting the work of researchers in the confirmation process. Once test results images have been input, the app can automatically make judgements on around 1,000 images in just a few minutes. As a result, researchers can now focus on new experiments or other tasks that require more human resources.

What did you pay particular attention to when using the agile development method?

Koyama: We collected feedback from researchers regarding their requests and the usability of the app, for example, and executed rapid sprint cycles in which we implemented and improved the required features.

 

In the conventional waterfall model, a significant amount of time is spent on elaborate requirements definition, and it is only when these requirements have been confirmed that planning and implementation can begin. As such, even if new requirements arise in the implementation process, sometimes they cannot be reflected into the system’s specifications. Moreover, when outsourcing development, changes to requirements lead to additional orders and the potential for additional costs. Agile development, on the other hand, assumes that plans and circumstances will change. One method within agile development is the scrum framework. In this framework, development is conducted within a two to three week cycle known as a sprint, and after the cycle is complete a usable app is released to the user. An actual functioning app makes it easier for the user to provide feedback, and the engineers can then use this feedback to make improvements in the next sprint.

 

As a specific example, at the start of the project we created a user interface in which users would use a computer mouse to select images. We then received an additional request from a user who wanted to be able to select images using the keyboard. While it might seem like a minor thing, for the researchers who actually use the app these detailed specifications can have a major impact on work efficiency. It was only because we quickly provided an actual app that we were able to receive such feedback so quickly.

 

How did the agile development method impact the project?

 

Goto: While we had previous experience of creating an in-house app as an extension of our analytical work, it was a real challenge to spend time creating such a complex interface using our capabilities alone. This is why I think the help of the Tech Workshop team played a major role in implementing the app in the business.

Quickly incorporating requests from users and making improvements allowed us to create an easy-to-use app in a short space of time.

 

Ukegawa: At the time, I was one of the researchers who used the app. While I wasn’t aware that an agile development method had been adopted in the project, I was surprised at just how quick in-house development was. We would provide feedback, and the improvements would be made just a month or two later.

What did you pay attention to as you worked alongside departments with different specialties?

Koyama: We focused on a thorough understanding of processes associated with the confirmation of crystallization and an understanding of the user’s psychology. To do so, we made use of the service design method to identify the operational processes and challenges. We broke down each process in detail and carefully listened to the thoughts of the researchers to clarify the specific issues they were facing. Enhancing our understanding of the operational processes enabled us to minimize any gaps between user expectations and the app’s actual capabilities and provide a product with real value.

Tanida: It took a long time to identify the operational processes, but it provided huge value. As researchers, to be honest I don’t think we thought that the developers would work to understand our operations so thoroughly. But through the process, we were able to accurately share the latent issues in research labs and procedures and clarify where the challenges lie. Because the developers were fellow employees it was easier to communicate to them the conditions of our experiments and our real thoughts.

 

Goto: Although in certain areas we worked with external partners on the app’s development, the support from the Tech Workshop in-house development team was hugely helpful. As the Tech Workshop is so well-versed in operational processes, challenges, and digital technologies, it played an important role in dealings with the external partners. The Tech Workshop found a common language, functioning as our interpreter and advisor to ensure we could proceed with the project with peace of mind. As a result, we succeeded in smooth communication and efficient development.

Agile development is highly regarded by users, but did you come up against any challenges in the project?

Koyama: The project made me realize that there are still some challenges to overcome if we are to work at the speed required by the business side of things. In particular, conventional in-house standards for the IT development process are based on the waterfall model, and so I think there is a need to formulate processes that are more suited to agile development. For example, we must simplify the approval process for minor system improvements. Outside of the narrow sense of development, such as design development, there are still factors that can hinder our agility. It will be important to eliminate these one by one.

Thank you. To finish, what are your expectations for agile development in pharmaceutical R&D moving forward?

Tanida: There are many patients across the world who are waiting for pharmaceuticals that do not yet exist. A single day’s delay in pharmaceutical development can prevent a patient from accessing the medicine they require. Through this project, I realized that compared to the conventional waterfall model, agile development allowed us to optimize the app by working at greater speed and by flexibly incorporating user feedback. Using this experience, I will work on R&D to ensure we can deliver the required pharmaceuticals to patients as quickly as possible.

 

Goto: Not limiting ourselves to this crystallization confirmation process, in the future we will aim to further enhance efficiency and gain new insights. To optimize the pharmaceutical production process and quickly deliver medicine to patients in need, we will continue to incorporate new technologies and methods and endeavor to improve our productivity.

 

Ukegawa: Our job is not only to produce pharmaceuticals but to reliably deliver them to all areas in need. If we only produce enough medicine for 100 people, we can only deliver it to 100 patients. Crystallization is one of the central process in the mass production of pharmaceuticals and this project has helped to improve the efficiency of the process. We will continue to work hard every day knowing that we have an important responsibility to look after patients’ lives.

 

Koyama: The project helped me to realize the high level of compatibility between Chugai Pharmaceutical’s research culture, which seeks to explore uncharted areas through trial and error, and the agile development approach, which aims to develop unprecedented products by constantly adapting to change.

In the future, we will continue to promote DX in our areas of R&D and co-create applications with real value for our researchers by utilizing our strengths at our agile in-house development. In doing so, we hope to help deliver even better pharmaceuticals to patients faster than ever.

 

Profiles

Tomotsugu Tanida (far right): Section Chief, Promotion Group 2, Pharmaceutical Planning and Promotion Department 
Product Owner of the project. Currently in charge of project management for a series of process development projects (CMC development) in pharmaceutical development, including drug substance research, drug product research, quality assurance, and quality control

 

Tomoya Ukegawa (far left): Formulation Fundamental Technology Group, Formulation Development Department

In charge of pharmaceutical pre-formulation research, with a particular focus on crystallization and physical property assessments. Current Product Owner of the project

 

Toshiki Goto (second from right): Bioinformatics Group, Biological Technology Department 
A data scientist in charge of next-generation sequencer analysis and image analysis. Oversaw the implementation of image recognition technology for the crystallization confirmation app.

 

Kenichi Koyama (second from left): Agile Development Group, Digital Strategy Department & Data Engineering Group, Modality Technology Department
 Oversees the planning and promotion of agile software development projects in the Research Division, while concurrently working for the in-house application development team named “tech-kōbō”.