Introduction

This article was originally published on our company’s note（Japanese only） in October 2024. To reach a wider audience, we are also reposting it here on our official tech blog. The content reflects the information available at the time of publication.
(All organizational names and job titles mentioned in the article are as of the time of the interview.)

---

ーーFollowing our first interview with our CTO, Mr. Nakamura, this time we are interviewing Mr. Isozaki, our Principal Investigator (PI). Thank you very much for joining us today, Mr. Isozaki.
To begin, could you tell us how you came to join digzyme? During your student days, you moved from the Tokyo Institute of Technology to the doctoral program at Kyoto University, correct? And after that, you joined digzyme.

Yes. I actually left the doctoral program at Kyoto University before completion. When I returned to Tokyo afterward, I found out that our CEO, Naoki Watarai — who had been my senior by one year at Tokyo Tech — had just founded digzyme. I thought it sounded really interesting, so I reached out to him, and that was the beginning.

ーーBack in your Tokyo Tech days, what kind of relationship did you have with our CEO, Naoki Watarai?

I originally met Naoki Watarai back at Tokyo Tech, when we were both involved in iGEM.（※The International Genetically Engineered Machine Competition）

ーーiGEM！digzyme is also supporting it as a sponsor.

Yes. iGEM is the world’s largest synthetic biology competition, where even undergraduates can engage in realistic research activities without joining a lab. During iGEM, Watarai-san and I were on the same team representing Tokyo Tech. It was a small team of just three people, so we got along quite well from the start. I could already tell back then that Watarai-san was an incredibly powerful and remarkable person. Over time, I became convinced that 'if Watarai-san starts a company, it’s going to be something really interesting.'

At the time, my own field of study didn’t give me deep expertise in enzymes, so I wasn’t that knowledgeable about them yet. After returning to Tokyo, I spent a while working in technical sales at another company. But after a short period of side work, I eventually joined digzyme full-time.

In my technical sales role, I would explain the platform we were handling to customers, then bring the information back to the company and hand it over to the appropriate engineers. That itself is an important job, and in fact it’s a position that digzyme still needs. However, most of the work was focused on customer negotiations, and I had very little direct involvement with new technical developments. I sometimes felt a sense of powerlessness in that kind of bridging role.

That experience was very valuable in giving me a hands-on understanding of what technical sales is like. But as a researcher, I believe creativity comes from being able to set your own challenges and run toward your own goals. At digzyme, I felt I could do just that, which is why I decided to commit fully. And I’m really happy that I can actually work this way now. I get a great sense of accomplishment tackling daily challenges.

ーーI see. It’s great to hear that you’re moving in the direction you want, and I feel happy listening to that! Specifically, what kinds of tasks or aspects do you find particularly rewarding?

I develop analytical technologies every day to solve challenges in collaborative research and new business projects. What I find particularly rewarding is the part where we discuss how to effectively translate biological features into dry analysis and then actually test them in practice. Specifically, we take the 'abstract challenges' we receive from our clients and break them down into very detailed hypotheses about the biological characteristics — testing them in dry analysis, and eventually connecting them to wet experiments. We refer to this process as 'translating' the features.

The procedure usually starts by identifying which aspects of biological features are likely to contribute to the problem we want to solve. In dry analysis, we formulate hypotheses and analyze things like structural differences, patterns in temporary sequence motifs, or whether the protein can be expressed in a particular species. But it doesn’t stop there: we also validate these hypotheses in the lab through wet experiments, or, in the case of collaborative research, have our partners test them directly.

Whether it’s collaborative research or our own projects, we can perform the full cycle of hypothesis testing — in other words, we can actually prove the hypotheses we’ve worked hard to develop. That is what gives me the greatest sense of fulfillment.

At digzyme, DRY and WET are seamlessly connected. Because these two foundations are solidly in place, it makes for a really excellent environment.

ーーNow, Isozaki-san, you are well-versed in both DRY and WET approaches… In your daily work, are there moments when you feel that your past experiences have been particularly helpful?

Actually, when it comes to DRY, my experience as a student was really just as a user — I could use some commercial analysis tools a little bit. I wasn’t able to write code or anything like that…

ーーWow! I can’t believe it…!

That’s right. And on top of that, I didn’t really know the logic behind how DRY tools are made — I was really just ‘able to use them.’ So I don’t think you could have called me a ‘DRY person’ back then (laughs). I was mainly a WET person.

During my doctoral program, I went offshore and studied marine bacteria. My main work was WET, and I only used Linux a little as a supplementary tool.

I used techniques like RNA-Seq — one of the next-generation sequencing (NGS) methods, which allows comprehensive detection of gene expression, known as the 'transcriptome' — to estimate the infection targets of marine viruses based on their genomic genes.

ーーI see. How about during your master’s program?

Up until my master’s program, I didn’t even know Linux, and I only used Windows — basically just Excel (laughs). At that time, I was even further away from genomics. During my master’s, I worked on cancer models using mice: transplanting cancer cells into mice, collecting samples, preparing tumor sections — it was almost like craftsmanship — and observing them under the microscope. You could say it was somewhat similar to what a pathologist does with human samples.

However, as I started attending academic conferences, I realized that if I couldn’t do any DRY work at all, I wouldn’t be able to survive, whether in academia or in any other research role.

I think many people who specialize in WET share this feeling. Most of the current members at digzyme would probably say the same thing: 'If possible, I really want to learn this too.

ーーBeing able to understand both DRY and WET approaches seems to have a lot of advantages.

Yes. But often, it can feel like you’re just pending, not really sure what to do. That’s part of the reason I deliberately changed labs when moving from my master’s to my doctoral program. I moved to a lab where I could potentially work with both WET and DRY. That said, as I mentioned earlier, I was only using commercial tools at the time, so I hadn’t even touched Python in terms of coding until I joined digzyme.

At digzyme, I learned everything from Watarai-san and Nakamura-san. Back then, we would occasionally hold programming sessions at a certain family restaurant (laughs), where they’d give me homework and I’d gradually improve.

Interestingly, having WET experience actually benefits DRY development. Understanding how WET researchers feel when using DRY tools is really valuable. Since we’re here, I can also explain what it means to do something in DRY that’s normally done in WET. First of all, you obviously can’t perform the experimental part of WET in DRY. What you can do in DRY is the investigative part. For example, at digzyme, 'enzyme discovery' is exactly that. Deciding which sequences to propose for a given challenge is normally part of the broader 'researcher' role. But because digzyme has the two foundations of DRY and WET, this 'enzyme discovery' is handled by DRY specialists.

If you can automate and quickly judge what sequences are likely to be effective in DRY, experiments in WET can proceed immediately afterward, and the number of experiments can also be reduced — there are many benefits. The same applies to 'enzyme modification.' You can quickly see in DRY where modifications should be made. But originally, all of this is part of the broader 'researcher' role. So, having WET experience naturally contributes to DRY development.

ーーGot it! That makes sense, thanks.
Going back a bit, could you tell us more about your research on marine viruses during your PhD?

I’ve always wanted to study viruses that exist in unusual places, not just in the ocean. As I mentioned earlier, I chose a lab where I could also do some DRY work, which led me to the Marine Molecular Microbiology Lab.

The motivation for wanting to study viruses in unusual environments goes back to when I was in elementary school. At that time, there was a somewhat popular book called The FUTURE is WILD (by Dougal Dixon and John Adams, with contributions by Takayuki Matsui and Akiko Tsuchiya), which simulated what kinds of creatures might thrive on Earth after humanity is gone. The book included CGI depictions of fantastical future creatures, like giant squid that looked almost like elephants — it was really captivating.

I was also interested in paleontology books. In evolution, when you trace ancient organisms, they generally get smaller — starting from microbes and gradually getting bigger is quite different from the world we live in now. I found that monster-like, imaginative aspect really fun (laughs).

So, I could have gone into a field like 'extremophile microbiology,' but I was drawn to something a bit more unusual. For example, there’s a hypothesis that viruses might have contributed to the development of mammals. The genes responsible for forming the placenta are composed of retroviruses, and in fact, the reason this suddenly appeared can be explained by viral evolution. Studying particularly unusual viruses like these seemed fascinating, and that’s what influenced my choice of research direction.

ーーI’m also curious about what viruses really are — they’re quite fascinating. Now, changing the topic a bit, could you tell us about any challenges you’ve faced in your work so far?

The main challenge is translating the abstract requests from clients into 'analytical approaches that are both commercially valuable and feasible.' Often, clients don’t know how detailed their requests should be for digzyme, or they simply can’t specify the level of detail, which makes the requests quite abstract.

For example, they might say something like, 'We don’t even have a concrete chemical structure yet, but we want something with roughly these properties,' or, 'It would be great if the production of this current product could be increased — we’re not sure whether an enzyme can be used or not.' When addressing these kinds of requests, we have to compare conventional methods and competitors with digzyme’s unique technologies, assess whether we truly have an advantage, and also consider market fit and other perspectives. Turning all of this into a concrete analytical approach is a challenge each time.

As for how we overcome this, we resolve it by consulting the right people, both inside and outside the company, as early as possible. Also, now that the business team has grown, I’m less often involved in the initial client meetings, so I personally encounter these abstract requests less frequently. I’m grateful that, as a company, the layer that converts abstract challenges into concrete plans has been steadily strengthened.

ーーWith the team growing, it seems that the environment for solving challenges is steadily improving. By the way, speaking of 'consulting the right people inside and outside the company,' Isozaki-san, I get the impression that you communicate very actively within the company. I personally benefit from that all the time, but do you do the same externally as well?

Yes. For new business projects, I often interview experts using platforms like VisasQ. In particular, for topics related to the Food Business Division, we sometimes gather experts through Mr. Takuo Miyauchi (Director and CSMO) and get their insights.

ーーThat’s reassuring. Are there any insights from these interviews that have particularly stuck with you?

There are many, but as one example… When we asked whether it would be commercially viable if the activity of a certain enzyme (the kind used in tests) could be greatly increased, we received the following comment:

‘The activity of this enzyme has never increased beyond a certain point. Of course, if it could be increased, the accuracy of the tests would improve, which would be good… but I’ve never seen such a case from a technical standpoint.’

ーー‘I’ve never seen anything like that!’ — That must have been really thrilling!

Yes. I could confidently tell myself, ‘This will really hit the market.

ーーThank you for sharing such a wonderful example. I have a follow-up question: what principles or values do you prioritize in your work?

For me, it’s about keeping open communication both inside and outside the company. To achieve this, I focus on building relationships where it’s easy to consult each other early when problems arise.

When psychological barriers are low, it’s less likely for relationships to turn sour. I’m careful not to signal any ‘walls’ to others, because visible barriers can create an ‘us versus them’ mindset. As long as the relationship is at least neutral, it prevents unnecessary conflict and makes problems less likely to occur.

ーーThat’s quite moving. Speaking of Isozaki-san, what kind of challenges would you like to take on in the future?

What I am currently working on is the overall management of WET research operations.
In the past, employees were organized flatly under our CTO, Nakamura. However, starting this year, Takayama-san (PI = Principal Investigator, Hiroo Takayama) and I have taken on a layer above all WET members as PIs. Under us are the PMs (Project Managers), while DRY members report directly to Nakamura.

Since I’ve never directly managed the research work of WET members before, it has been a bit challenging. I am moving forward with support from Takayama-san, my fellow PI.

Looking ahead, I hope digzyme will strengthen its PM layer even further. Ideally, we want to increase the number of WET members who have some understanding of upstream (DRY) work.

That said, I realize that diving straight into DRY work can be confusing—people might not know exactly what to learn at first. That’s why I plan to convey this gradually through on-the-job training (OJT).

ーーI see. If the number of WET members capable of DRY analysis continues to grow, the organization will become even stronger. Are there any other challenges you would like to take on?

I want to promote the development of analysis pipelines that reflect the characteristics of enzyme families based on their three-dimensional structures. By doing so, we can tap into unmet needs across the enzyme industry and explore novel approaches that haven’t been tried before, which should lead to greater diversification and improved productivity in enzyme-based manufacturing. I hope to advance this together with Mr. Koichi Tamura, our informatics specialist.

digzyme’s strength lies in our ability to handle both upstream and downstream processes, while also preparing for production. This is fundamentally different from the nuance of “AI doing a bit of analysis for you.

ーーIt seems like this will become a move that truly leverages our strengths.

Yes. I think digzyme’s biggest distinguishing feature is that it’s deep tech with a genuine market-in mindset. In particular, all the executives naturally think in terms of scaling.
—Of course, the fact that we can differentiate ourselves in specific technologies comes from having professionals from various fields gathered here—but even putting aside whether it’s technically feasible right now, we have our antennas up for ideas that could significantly propel the business forward. That’s why we have this virtuous cycle: we can define the necessary technologies and reliably develop them.

In some cases, deep tech companies aren’t focused on scaling. But our philosophy is that a vague notion like “it’s fine as long as it makes some money” or “this technology is amazing, so surely everyone will want it” just won’t cut it. No matter how impressive a technology is, if it’s for applications people don’t actually want, it can easily end with a “Sure, it’s innovative, but…” and go nowhere.

Because we have a market-in mindset, we can truly move our customers. That’s why it’s so important to have people like Mr. Watarai leading the charge—investigating the cutting edge of business with scaling in mind, and translating those insights into precise technical language.

ーーFinally, could you share a few words for those future colleagues who are considering applying to digzyme?

We have a diverse team in terms of both personality and research or professional backgrounds, offering opportunities to gain experience across a wide range of fields beyond just research. So, if you’re eager to take on new challenges, you’re very welcome here!

▼Original Post［note（Japanese only）］
https://note.com/digzyme/n/n4cb24197110b

Introduction

This article is a reprint of an interview originally published on our company’s note page in September 2024. To make it accessible to a wider audience, we are sharing it here on our official tech blog. Please note that the content reflects the context and information available at the time of the original publication.

---

— This marks the very first interview in our series! Thank you for joining us, Nakamura-san.
To begin, could you tell us how you joined the company? …Although, come to think of it, you’re actually one of the founding members, aren’t you?

Yes, that’s right.
I co-founded digzyme Inc. while I was still a student at Tokyo Institute of Technology, together with Mr. Torai¹, who is now our CEO, and Professor Yamada², who serves as our CSO and is an associate professor at Tokyo Tech.

The inspiration for starting the company came from research we were conducting in the Yamada Lab³, which later became the foundation of digzyme. We saw strong potential for turning that research into a viable business.

— The research that became the foundation of the company… we’d love to hear more about it. What aspects made you see its potential as a business?

It was research that eventually evolved into what is now digzyme Moonlight™, focused on enzyme discovery.

At the time, Professor Yamada and Nagase & Co., Ltd. were conducting a joint research project. I joined them and contributed by performing additional computational analyses.

Specifically, Nagase approached us with a request:
"We’re looking for an enzyme with these specific properties."
Our task was to search for suitable enzymes from a database based on that request. The process was going well—we were consistently identifying promising candidates. So, the initial plan was to publish the results as a research paper.

From there, I started thinking: What if this discovery technology could be applied to other domains as well?
That line of thinking gradually led us to explore broader possibilities—and that was the starting point for digzyme’s business model.

In terms of its broader applications, what I specifically had in mind was this:
Could we identify enzymes capable of catalyzing synthetic reactions for compounds where no such enzymes have been discovered yet?

That’s exactly what we achieved through our collaboration with Nagase.

The project focused on enzymes from a particular plant that produces a unique compound. But extracting the enzyme by physically grinding the plant each time? That’s obviously not scalable.

So we thought: What if we could reproduce this biosynthesis using microbial enzymes instead?
We began screening for enzymes that might be able to catalyze the same reaction—and we found quite a few.

In fact, we identified a number of enzymes that could potentially synthesize previously unknown or commercially unavailable compounds. That was really exciting.

— That’s incredible!

Yes, and from there, we began to discuss in the paper the possibility that these enzymes might allow us to synthesize compounds that were previously considered extremely difficult to produce—with much more ease than expected.

When I reported this to Professor Yamada, he pointed out that there must be many people out there who need exactly these kinds of enzymes. That led to the idea that this could be really interesting as a business.

And the timing couldn’t have been better—around that same time, Mr. Torai was actively exploring the idea of launching a startup.
So with that momentum, the three of us—Professor Yamada, Mr. Torai, and myself—decided to start working together.

Once we did, we immediately began receiving inquiries from people saying,
"If there’s a technology that can identify enzymes like that, we’d love to use it."
That strong interest convinced us: there’s real demand out there—so we decided to found the company.

ーー Indeed, conventional enzyme development has required an enormous amount of trial and error to identify genes encoding enzymes suited to specific purposes.
Because of this, there has been a heavy reliance on serendipitous discoveries, which brings uncertainty and results in enormous development costs. Considering these challenges, it’s clear there is substantial demand for better methods.
After founding the company, what kinds of work have you been involved in?

Right after the company was founded, I worked on developing the core technology — the enzyme discovery software. At the same time, we began exploring how to further apply this technology to develop areas that would become key strengths for the company.

We already had several research and development themes internally, so I was also involved in those development projects.

However, I was still a student for the first six months after founding the company. After graduation, I worked at a pharmaceutical company for nearly two years.
During that period, I participated in discussions around technology development while balancing this work alongside my role at digzyme.

— After that, you transitioned to committing full-time to digzyme, correct?

Yes. Around the time I fully committed to digzyme, a project funded by NEDO⁴ (New Energy and Industrial Technology Development Organization) started.

In this project, we worked on designing spotlight⁵ (digzyme Spotlight™, our enzyme function improvement platform). Together with Mr. Torai and the team, we discussed ideas such as:
“If we build a machine learning model like this, it should work well.”

Once the budget was secured, we allocated resources to the researchers accordingly, directing who should develop which parts, and proceeded with the development in that manner.

— Naturally, you gradually took on the role of leader for each project. You also became responsible for recruitment and training as CTO, correct?

Yes. As the team gradually grew, I also began training new employees in dry lab research techniques.
The very first was Mr. Isozaki⁶, who joined us as a part-time staff member just after the company was founded. Since he originally came from a wet lab background rather than dry lab, Mr. Torai and I taught him the dry lab techniques.

— I see. So how do you actually go about training people in dry lab techniques?

Basically, I believe the best way is to work through real examples.
When people study programming in general, they often start with simple exercises like “what happens if you program 1 + 1 = 2?” But…
that kind of practice doesn’t really stick, and it often leaves people wondering, “How does this actually apply in real life?”

— That makes sense.

If you don’t tackle real-world problems, it’s often not very engaging and doesn’t translate well into practical skills. So I think assigning realistic tasks is crucial for effective learning.
For example, the training materials I create include problems I personally worked on in the past. Trainees solve these problems and go through the entire process as a way to learn.
More recently, we often use actual challenges from our clients as test cases, and have trainees work through them together with Mr. Takayama⁷ and Mr. Isozaki.

— Thank you for the detailed explanation. How have your recent responsibilities been?ですか？

My main responsibilities have been managing individual projects and overseeing research resource allocation.

— How do you feel about taking on management responsibilities?

For me, management has never really been something I struggled with. Even when I was a student working part-time, I often took on roles like team leader, managing members and coordinating tasks. I have this feeling that I can’t be satisfied unless I’m involved in the core aspects of things (laughs).

When I worked at the pharmaceutical company, I wasn’t the type to be content just doing experiments or research at the ground level—I needed to dive into the detailed discussions and the essence of the projects to feel satisfied.
Of course, I enjoy the research itself and want to keep doing it, but I’m not someone who can just work blindly without fully understanding what’s going on…
In that sense, I guess naturally I ended up taking on management roles.

That said, I’ve been gradually handing over these management responsibilities to Mr. Isozaki and Mr. Takayama, and I’m now returning more to my core work—leading new technology development and building foundational technologies.
Of course, I’m always thinking about what kind of new technologies our dry and wet lab teams should have, and brainstorming ideas to further strengthen digzyme’s competitive advantages.

— I’m very interested in the ideas for further strengthening digzyme’s competitive advantages.

Absolutely. Regarding how we can grow going forward, I want us to thoroughly discuss and develop the necessary software and wet lab technologies in the business areas that the divisions have identified as targets for expansion.

At the same time, I personally focus on ensuring that the foundational tools and platforms we build within each project are effectively utilized to drive the progress of those projects.

— Nakamura-san, what aspects of working at digzyme do you find most rewarding?

I find it very rewarding to apply new technologies to real-world challenges, developing and updating the missing pieces as we go.

For example, in developing digzyme Spotlight™, we were pioneering the use of AI and machine learning to improve enzymes at a time when almost no one else in the world was doing this.

Before that, it was common to study enzyme structures and conduct research like,
"If we change the part of the protein that interacts with the substrate, the substrate might be affected as well, so let’s mutate it to improve activity."
However, AI and machine learning methods were still not widely adopted.

At the same time, there were many requests to improve enzyme performance—it was a challenge across the entire enzyme industry.
I enjoy facing challenges and solving problems, so I found it very rewarding to create and develop programs while discussing ideas like,
"With digzyme, we should be able to predict and design mutants that increase activity using AI."

But honestly, I enjoy tackling problems of any size, whether big or small.
The Spotlight project was somewhat large scale, but even fixing small, everyday annoyances—things like "this is kind of a hassle"—and watching the system run more smoothly is something I really like.
No matter the scale, I find great satisfaction in new technology development and problem-solving—the process of continuous improvement.

— I see. It sounds like digzyme is really supported by you, Nakamura-san, who finds fulfillment in solving challenges big and small! Since we’re on this topic, I’d love to hear more about the uniqueness of Spotlight.
I’ve heard that the platform was developed by a team that fully leveraged each member’s unique background. Could you please tell us more about this aspect?

Spotlight is a program that uses machine learning algorithms to predict
“If we do it this way, we should be able to identify which parts of the enzyme to modify.”

I had been studying machine learning throughout my student days and also while working at a pharmaceutical company, so I was able to apply that knowledge to create the platform.

Regarding members who are experts in sequence analysis, there’s Mr. Torai and Mr. Hikoyuu⁸ (Informatics Specialist Mr. Hikoyuu Suzuki). They have been studying genome-level gene and protein sequence analysis extensively in the lab.

Then there’s Mr. Tamura⁹ (Informatics Specialist Mr. Koichi Tamura), who is highly knowledgeable about three-dimensional structural data.

So the three of them—Torai, Hikoyuu, and Tamura—worked on determining which features the model should learn from, combining sequence and structural data expertise. Meanwhile, I focused on conceptualizing the machine learning models and approaches.

Finally, Mr. Isozaki implemented the system, and that’s how Spotlight was completed.

— It’s truly moving to hear how you’ve brought together such collective wisdom. May I ask about the challenges you’ve faced at work, and what helped you overcome them?

Rather than challenges per se, I’d say that recruitment has been quite tough.
It’s a significant matter both for the company and for the candidates whose lives are deeply impacted, so I recognize it as a serious responsibility.

In that context, I struggled a lot with how to make the right decisions when hiring people who will shape digzyme’s future.
After several rounds of recruitment, I feel like I’ve finally gotten the hang of it.
Mr. Torai has a very good sense of how to conduct interviews and ask questions, so I’ve learned a lot by following his example.

As for what kind of people we specifically look for, it definitely comes down to those who don’t give off any sense of incongruity during conversations.
We want candidates who not only respond within expected parameters but can also go beyond that in their answers—those are the people we want to hire. On the other hand, if their answers seem stuck one or two steps behind what we expect, it’s a bit difficult to move forward with them.

It’s also important that candidates are good at troubleshooting. Wet lab research especially comes with its share of failures.
In dry lab work, if something goes wrong, you can usually retry quickly—and that’s a field where I tend to come up with ideas easily. But with wet experiments, if you want to redo something, you might lose a whole week, causing significant schedule shifts.

Honestly, I’m not that well-versed in wet lab work myself, so when something goes wrong, it’s important for me to have someone who knows more than I do and can think and act independently during the problem-solving phase.
Troubleshooting experimental issues happens quite often, so we try to hire people who can handle these situations well.
We ask candidates how they’ve handled failures in the past to assess their troubleshooting skills and make sure we bring in capable individuals.

— I see. Since you mentioned WET lab, it makes sense that digzyme’s WET capabilities are so impressive given the team you’ve built.

Exactly. I believe one of digzyme’s strengths in WET lab is that our team can handle a surprisingly wide range of tasks.
For example, when we want to evaluate a certain enzyme, we read relevant papers, develop protocols, try experiments, express proteins, and perform the evaluation. Of course, this requires researchers who can actually conduct solid scientific work.
It’s not something anyone can just casually do by saying, “Hey, read this paper and try replicating the experiment!” (laughs)
Being able to handle that naturally is actually a very high-level skill.

Conversely, it’s very rare that our dry lab analysis gets stalled because of issues on the wet lab side. I say this casually, but it’s actually a remarkable achievement.

That said, while our technical skills are very high, we’re not particularly strong in terms of resources. Compared to many companies and academic labs, we don’t possess special microbial strains or proprietary genetic engineering techniques. We primarily use publicly available materials.
So, honestly, we don’t have an edge in terms of resources, but I take pride in the strong abilities of our research staff.

— I see. That’s reassuring to hear.

Yes. By the way, in DRY lab work, we often don’t know the exact causal relationships. There are many uncertainties about which is the cause and which is the effect.
So when we analyze enzymes, we proceed while considering the possibility of false positives.
We narrow down candidates to very promising ones, but after that, we rely heavily on the high level of WET lab expertise.

For example, even when using E. coli, they don’t just use one strain—they prepare multiple strains, as well as various other organisms—and skillfully conduct experiments to overcome challenges. That expertise is invaluable.

— Looking ahead, what kinds of challenges would you like to take on?

Basically, I hope the projects we’re working on progress through their stages and eventually get launched as products that genuinely make a difference.
I’m excited about the enzymes we’re currently developing becoming actual products — it would be great to say, “This enzyme is actually in that product!” someday.

From a technical standpoint, as projects advance, new challenges unique to those stages will arise, and I want to tackle those.

For example, when you want to convert a certain compound into another, the numerical goals like “the amount of enzyme required” or “the efficiency needed” will become more concrete than they are now.
Achieving—or not achieving—those targets will be a crucial issue in the near future.

The next stage will be “mass production.” We will need to set production performance goals such as “to meet this product price, the culture medium must be this volume,” or “we need to produce this much enzyme.”
Since these goals directly impact business continuity, we must resolve them properly.

Also, I’d love to try developing completely artificial enzymes.

By “completely,” I mean—it’s a tough challenge (laughs). Usually, we base enzyme design on natural enzymes found in microbes or improve upon them. But now, with AI technology, it’s becoming possible to design enzymes from scratch, purely from data.
This means you could design enzymes on a computer that aren’t really based on any natural microbial enzyme, though they might bear some resemblance.

Of course, whether people would want to eat food containing such artificially designed enzymes is another question (laughs).

Currently, we’re bound by natural enzymes as the base, but this approach allows us to break free and create entirely original enzymes.
Even if practical use is uncertain, it’s exciting because it feels truly novel.

I’m also interested in developing systems that don’t rely on microbes, like cell-free systems.

We’re already exploring and discussing cell-free approaches internally, but basically, wet lab processes still mostly involve genetically modifying microbes to express proteins. Often, “protein expression fails,” which is a big hurdle.
Cell-free systems can sometimes reduce that problem—though expression failure can also happen there—so I want to try that.
In any case, I hope to eliminate the unique uncertainties of bio processes.

— When you say "the unique uncertainties of bio processes," what do you mean exactly?

In biological experiments, there are often cases where things just don’t work well, and nobody really knows why.
I think it would be amazing if that “I don’t really understand why it’s failing” part could be eliminated.

I’m not saying we already have a way to guarantee success—this is really just a dream at this point (laughs).

For example, take just the “culture conditions.”
There’s no theoretical way to know exactly which culture conditions are best. When culturing a microbe, you repeatedly try different combinations of ingredients in the medium and experimentally find which one works best.
It’s not like you can say, “This is the best condition” based purely on theory—it’s more like, “I don’t really know why, but this works better.”

Sometimes the culture grows well, and sometimes it doesn’t. Protein production is similar—it’s not consistent; sometimes you get a lot, other times less. There’s quite a bit of variation.

Biological experiments inherently have these fluctuations.

“Failure” is an extreme example of this uncertainty, but even when things are going well, there are times when it’s “especially good” and times when it’s “just okay,” so there’s a lot of variability and error.

Even amid these uncertainties, one major challenge remains “poor expression depending on the host organism.”
To improve this, we often think, “It would be great if there was a system that could express any enzyme from any organism.”
If such a system exists, I feel that cell-free systems might be the answer.

— Expression of anything—that’s quite a dream. Thank you for the detailed explanation. Lastly, do you have a message for future team members considering applying to digzyme?

“Let’s tackle the ambiguous challenges of biology together with cutting-edge technology and innovative ideas!”

— Thank you very much, Mr. Nakamura.

---

¹ Nao Torai – CEO and co-founder of digzyme Inc.
² Dr. Takuji Yamada – Associate Professor at Tokyo Institute of Technology and CSO of digzyme Inc.
³ Yamada Lab – Laboratory for Life Science and Technology at Tokyo Institute of Technology.
⁴ NEDO: Japan’s national agency for promoting research and development of new energy and industrial technologies.
⁵ spotlight: digzyme’s platform for enzyme function improvement through machine learning.
⁶ Principal Investigator Mr. Tatsuhiro Isozaki
⁷ Principal Investigator Mr. Yuki Takayama
⁸ Informatics Specialist Mr. Hikoyuu Suzuki
⁹ Informatics Specialist Mr. Koichi Tamura

Closing Remarks

▼ Original article is available here (note)
https://note.com/digzyme/n/n4cb24197110b

Introduction

From May 21 (Wed) to May 23 (Fri), 2025, ifia JAPAN 2025 was held over three days.
As with last year, our CEO, Dr. Watarai, gave an exhibitor presentation at the event.
The full presentation is now available on YouTube—please feel free to take a look.

In this exhibitor presentation, we introduced the newly launched “digzyme Custom Enzyme Lab,” unveiled on May 21, 2025.
The session covered two key technological approaches: DRY (bioinformatics-based analysis) and WET (experimental validation), and provided an overview of the entire platform.

This article takes a deeper dive into two potential real-world applications of the digzyme Custom Enzyme Lab, which were briefly mentioned during the presentation.
Through a Q&A format and from the perspective of our CEO Dr. Watarai, we explore the technical breakthroughs behind each case, as well as the in silico design strategies employed.

While the presentation offered a high-level overview, this article aims to give you a more concrete understanding of the capabilities and potential of the digzyme Custom Enzyme Lab.

We invite you to read on and explore the details—beginning with the first case study.

Expected Application Case 1 of the digzyme Custom Enzyme Lab

Q: What do you consider the most significant value of this result?
A: The physical properties of carbohydrates vary depending on the linkage patterns between constituent monosaccharides.
This case is particularly valuable because it represents a rare example—even in academic contexts—where in silico techniques successfully identified an enzyme capable of constructing a specific glycan structure.
Moreover, the target enzyme was discovered with just 10 experimental validations, which highlights the efficiency and precision of the approach.

Q: What was innovative about this approach compared to conventional methods?
A:In this case, our proprietary, detailed analytical techniques ultimately proved effective when applied to the deep learning (DL)-based structural prediction technologies of the time, such as AlphaFold2. Traditional homology-based models had difficulty predicting subtle structural differences in proteins that lead to variations in glycan structures. However, the AI technologies available at the time enabled us to capture some of these critical features to a certain extent.
(Note: As there is still a gap between these earlier AI technologies and today's cutting-edge generative models, we use the term "AI" here for convenience.)

Q: What team efforts or contributions led to this success?
A: The lead researcher deeply investigated the client’s specific needs and successfully translated them into tailored screening criteria for enzyme selection.
By working closely with our core development team, a customized analysis pipeline was developed, which was crucial to achieving this outcome.
We believe one of our key strengths is the ability to flexibly build new tools and solutions beyond our existing platforms to meet unique and complex challenges.

---

Next, let us introduce the second case study, which was conducted in collaboration with Mitsubishi Chemical Corporation.

Expected Application Case 2 of the digzyme Custom Enzyme Lab

Q: What do you consider the most significant value of this result?
A: PVC (polyvinyl chloride) is a synthetic compound whose mass production began in the 20th century and does not exist in nature.
Assuming that natural microorganisms have not evolved degradation mechanisms for such materials, it would be highly unlikely to discover well-optimized degrading enzymes from natural sources.
However, living organisms are known to retain a wide variety of “non-optimized” or dormant genes within their genomes, which may later contribute to adaptation under environmental pressure.
This case can be seen as an attempt to identify such latent enzymatic functions through in silico screening—making it a particularly challenging theme.

Q: How long would it have taken to discover such an enzyme using conventional methods?
A: In recent years, there have been several studies that identify artificial plastic-degrading enzymes using methods akin to enrichment culturing. For example, researchers may submerge a particular type of plastic resin in the seabed for an extended period, then retrieve and observe its degradation, or isolate and culture microbes from biofilms formed on the plastic.
When successful, these efforts can uncover microorganisms with plastic-degrading enzymes, allowing identification through genomic analysis or BAC library construction. However, due to the inherently slow degradation process, such approaches often require years to yield results.
Moreover, it is common for degradation not to occur at all, resulting in unsuccessful attempts. In contrast, in silico discovery can typically be completed within about six months, making it a relatively efficient method even for targets that would otherwise require long-term experimental work.

---

Conclusion

Reflecting on the presentation, Dr. Watarai shared the following comment:

“With digzyme Custom Enzyme Lab, we are able to prepare in silico libraries in advance—similar to what we did in these collaborative cases. It’s a service we recommend to customers seeking to test purified enzymes from high-precision candidate libraries.”

As this statement illustrates, a bioinformatics-based approach to enzyme design has the potential to dramatically accelerate practical enzyme development, even under resource-constrained conditions.
As applications continue to expand across diverse domains, digzyme Custom Enzyme Lab is expected to play a pivotal role as a core technological foundation.

Introduction

My name is Murase from the Food Business Division.
Our company exhibited at "ifia JAPAN 2025 – The 30th International Food Ingredients & Additives Exhibition and Conference", held at Tokyo Big Sight from Wednesday, May 21 to Friday, May 23, 2025, following our participation last year.

During the exhibition, we had the valuable opportunity to engage directly with many visitors who showed strong interest in our technologies.
At our booth, we introduced our latest initiatives to these attendees. One of the main highlights was the launch of our new solution, “digzyme Custom Enzyme Lab”
(For more details, please refer to our press release:https://prtimes.jp/main/html/rd/p/000000018.000050097.html）

The launch received an overwhelmingly positive response, far exceeding our expectations. Our booth was filled with lively discussions throughout the exhibition, as we received numerous specific questions and inquiries from many visitors each day.

In this special edition of our tech blog, commemorating the launch of “digzyme Custom Enzyme Lab”, we’ve selected some of the most frequently asked questions from the exhibition and provided detailed answers in a Q&A format.

This post is not only for those interested in our new solution, but also for anyone curious about enzyme-based development who may be wondering where to start.
We hope you’ll find useful insights—please read on to the end!

---

Q: For what types of product development can “digzyme Custom Enzyme Lab” be applied?

A：“digzyme Custom Enzyme Lab” is a flexible solution that can be applied to a wide range of development themes—from specific goals such as improving the efficiency of existing enzyme-based manufacturing processes to broader, more exploratory themes like developing novel food ingredients using enzymes.

By repeatedly exchanging purified enzyme samples and receiving feedback from your in-house evaluations, the development direction can be adjusted flexibly at each stage.

Q: What kind of information is provided with the purified enzyme samples?

A：We perform preliminary testing to confirm enzyme activity and provide a profile including optimal temperature, optimal pH, thermal stability, and pH stability. These data are provided alongside the purified enzyme samples.
Verification in your specific application or evaluation system can be conducted by your team.

Q：What is the quantity of purified enzyme included in the sample?

A：The quantity depends on the development theme and is determined through consultation. As a general guideline, samples are typically provided in volumes of several milliliters of enzyme solution, equivalent to several milligrams of protein.

Q：How do you define or set the initial development timeline?

A：Following a prior evaluation of the requested development theme, we assess the feasibility and propose an initial development timeline.
In most cases, the initial phase—covering in silico enzyme design through to the first delivery of a purified enzyme sample—is completed within 2 to 6 months.

Q：Is non-GMO enzyme development an option?

A：Yes, it is possible. For more details, please refer to the “digzyme Express” introduction page:https://www.digzyme.com/cms/wp-content/uploads/digzyme_Express_ol.pdf

Q：Is “digzyme Custom Enzyme Lab” a solution exclusively for the food industry?

A：“digzyme Custom Enzyme Lab” is a versatile solution available for use not only in the food industry but also in other sectors, including the chemical industry.

Q：If a suitable enzyme is found among the provided purified enzyme samples, what happens next?

A：Enzymes developed via “digzyme Custom Enzyme Lab” can smoothly transition into manufacturing development. digzyme provides comprehensive support throughout the entire process, including manufacturing technology development and regulatory approvals, accompanying you until your project is fully commercialized.

Q：How is intellectual property handled for the developed enzyme library?

A：If you find a promising enzyme among those developed via “digzyme Custom Enzyme Lab” and decide to pursue its commercialization, we are prepared to accommodate your needs flexibly.

---

This concludes our responses regarding the services provided through “digzyme Custom Enzyme Lab”.
Please feel free to contact us anytime, as we remain flexible and ready to accommodate your specific needs during the actual development process.

Thank you very much for reading through this Q&A.

If you have any questions or require further clarification, please do not hesitate to reach out to us via the contact form below.

[▼ Contact Form]
https://www.digzyme.com/contact/

Koichi Tamura
(Research and Development Department)

Introduction

Enzymes are biopolymers (mainly proteins) that catalyze a wide variety of chemical reactions. Due to their lower environmental impact and high selectivity compared to metal catalysts, enzymes are widely used in industrial, food, and other applications. In general, the rate of chemical reactions depends on temperature; as the temperature increases, the reaction rate also rises. This is because heating provides the energy needed to overcome the reaction's activation barrier. For example, the Haber-Bosch process, which synthesizes ammonia from nitrogen and hydrogen using inorganic iron-based catalysts, operates at temperatures approaching 500°C.

In contrast, enzymes, being biopolymers, cannot withstand such extreme conditions and unfold (denature) at a certain temperature. This temperature is referred to as the melting temperature (T_m). T_m is a critical indicator for assessing an enzyme's thermal stability, and predicting and improving T_m is a key challenge in protein engineering. At digzyme, we have developed a computational method to predict scores correlated with enzyme T_m, enabling the selection of enzymes that function at high temperatures and the improvement of enzyme performance to meet a broader range of needs.

In this blog, we evaluate the performance of our developed method using two datasets of experimentally derived protein thermal stability data.

Methodology

In the context of enzyme discovery and modification, target enzymes for thermal stability prediction often originate from a wide variety of species and protein families. Therefore, prediction models require versatility, independent of these two factors. To meet this requirement, various methods have been proposed, which can be broadly categorized into the following two approaches:

1. Data-Driven Approaches

This approach, often referred to as machine learning, involves constructing predictive models for protein thermal stability based on large datasets derived from experimental results. The process of building these models is called training. The duration of training depends on the amount of data and the complexity of the model, ranging from the time it takes to enjoy a cup of coffee to an entire week of computation using cutting-edge GPGPU (General-Purpose Graphics Processing Unit) systems. Once a trained model is established, the computational cost for subsequent predictions is significantly lower than during the training phase.

As suggested by the model construction process, the accuracy of machine learning models heavily depends on the quantity and quality of the training data. Limited data introduces bias into the inferred rules, and care must also be taken to ensure the training data is not skewed toward specific species or protein families. Additionally, consistent experimental conditions for measuring protein properties are desirable, although achieving this is challenging given that data in current databases originates from various research groups.

Even with attention to data quantity and bias, generalizability issues may persist, primarily due to overfitting and extrapolation challenges. Overfitting occurs when a machine learning model overly adapts to the training data, yielding excellent predictions for the training set but failing with new data. Preventing overfitting requires expert techniques, such as appropriately controlling model complexity. Extrapolation refers to the model's ability to maintain reliability for data outside the training range. For proteins, this means being able to make reliable predictions for novel amino acid sequences with low homology to the training data. This capability is critical in enzyme discovery and modification because the target sequences often show low similarity to previously studied enzymes. While extrapolation is a crucial feature, building models inherently equipped with it is difficult except for simple linear models. Thus, the best practice here is to thoroughly understand and respect the applicability domain of the constructed machine learning models.

2. Physics-Based Approaches

Physics-based models begin by constructing energy functions that describe atomic interactions. The universal natural laws governing atomic and molecular systems have been mathematically formulated for over a century. Simple examples allow manual solutions to these equations, enabling comparison with experimental values to validate their accuracy (see introductory physical chemistry textbooks for details). Since these equations include no arbitrary parameters apart from physical constants (e.g., the speed of light, Planck's constant), they offer universal applicability. However, for complex multi-atomic systems like enzymes in solvents, these equations become highly intricate, making their practical computation infeasible even with state-of-the-art computational resources.

To address this, approximations are introduced to simplify the equations, trading off precision and universality for computational feasibility. Common approximation methods include:

●Empirical Energy Functions in Molecular Mechanics: These functions simplify inherently complex 　atomic interactions by assuming basic functional forms and introducing parameter sets.
●Treatment of Solvents via Statistical Mechanics: Instead of explicitly modeling numerous solvent molecules around the enzyme, their effects are replaced with averaged quantities.

These approximations significantly reduce computational costs but at the expense of accuracy and universality.

digzyme Score

At digzyme, we adopt a physics-based approach to predict enzyme thermal stability. Our model takes three-dimensional structural information of enzymes as input and outputs a score—referred to as the "digzyme score." This score is designed to correlate with the enzyme's melting temperature (T_m) and typically falls near a value of 1.0. While it does not directly predict the exact T_m value, this design suffices for practical purposes, such as comparing the relative stability of multiple enzymes.

Case Study 1: Prediction of Mutant Thermal Stability

In order to computationally design and select useful enzyme mutants, it is necessary to calculate various properties of numerous candidate mutants and rank them based on their values. Among these properties, thermal stability is particularly important. Below, we compare the results of multiple groups and digzyme in a competition to predict the difference in unfolding free energy (ΔΔG_u) between the wild-type and mutants of frataxin, an enzyme involved in iron-sulfur cluster regeneration.

A competition was held in 2018 to predict the difference in unfolding free energy (ΔΔG_u) between the wild-type and eight mutants of frataxin[1], and the results were published in a paper in 2019[2]. Unfolding free energy (ΔG_u) is defined by the equation (1), where...

ΔG_u = G_u - G_f                                                              (1)

It is defined as the change in free energy associated with the unfolding of the enzyme. Here, G_u and G_f  represent the free energies of the unfolded and folded states, respectively. In general, since the folded state of the enzyme is more stable, ΔG_u is greater than 0. Note that the larger the ΔG_u, the more difficult it is for the enzyme to unfold (i.e., the more stable it is). For both the wild-type and mutant forms, the unfolding free energy defined by equation (1) is measured, and by calculating the difference between them, the change in unfolding free energy due to the mutation can be computed as follows:

ΔΔG_u = ΔG_u(mutant) - ΔG_u(wild-type)                                       (2)

ΔΔG_u < 0 Then, it means that the enzyme has become destabilized due to the mutation.

Figure 1 on the left shows the predicted score ("digzyme score") calculated by digzyme and the experimental values. The Pearson correlation coefficient was 0.87. This result is slightly better than the existing popular physics-based method, FoldX (Figure 1 on the right).

Another physics-based method was used by the Pal Lab group, which employed a molecular dynamics (MD)-based approach for prediction. In this method, structural sampling (1 ns) is performed using MD, and the sampled structures are clustered into folded and unfolded states. Subsequently, energy calculations are performed for each representative structure in the clusters, and the values from equation (1) are calculated. By running this calculation for both the wild type and mutant, values can be obtained that can be compared with experimental data using equation (2). However, the assumption that the protein in the folded state (MD starting structure) will undergo a structural transition to the unfolded state within 1 ns under the given calculation conditions (27°C, typical equilibrium MD) is clearly unrealistic. This is because even for fast-folding proteins, which fold/unfold relatively quickly, simulations near their T_m still take over 1,000 times longer for a structural transition to occurr[3], the moderate correlation found using this method (Figure 1 on the right) is arguably a result of chance.

The group that proposed the model with the highest correlation in the contest was the Kim Lab group, with an absolute correlation coefficient of 0.89 (Figure 1, right). This group used a machine learning model for predictions, utilizing thermal stability data of mutants registered in the Protherm database for training [4]. In this machine learning model, both structure-based and amino acid sequence-based features are calculated, and a regression model is built using gradient boosting trees with these features. Among the structure-based features, the calculated values from the physical model FoldX are included, and it is known that these values are the most important features [4]. This fact suggests that using the digzyme score, a physical model with similar accuracy to that of FoldX, could help build even more advanced machine learning models. If there is a large amount of mutant data available for the target enzyme, it may be worth attempting to build a dedicated model.

Case 2. Prediction of Thermal Stability of Enzymes with the Same Function

In the search for useful enzymes, a large number of amino acid sequences of enzymes with the same (predicted) function are extracted from a database (forming the population), and candidate enzymes are selected by ranking them based on some criteria. Similar to mutant design, thermal stability is one of the important indicators in this process.

It is important to note that the prediction of thermal stability here is different from the prediction of the difference between wild-type and mutant enzymes as seen in Case 1. Typically, the length of the amino acid sequence is the same for both wild-type and mutant enzymes, and the sequence identity is often close to 99%. In contrast, in a population of sequences extracted from a database for a specific function, the length of the amino acid sequences varies, and the sequence identity is generally low. Predicting thermal stability and creating a ranking of sequences in such a population can often be challenging, as shown below.

The example here is a large-scale dataset of the T_m (melting temperature) of nanobodies (small fragments of antibodies) called NanoMelt, which was posted on a preprint server in September 2024[5]. This dataset includes both existing data and newly measured data by the authors (640 data points), with experimental conditions such as protein concentration, pH, and buffer standardized. For this dataset, the digzyme score was calculated for each enzyme using the same physical model as in Case 1, and the correlation with experimental values was examined.

As shown in Figure 2, except for the results where the NanoMelt dataset itself was used for training (Figure 2, lower right), there was no or only weak correlation between the experimental values and the prediction results. First, FoldX, a physical model that achieved a high correlation coefficient in Case 1, failed to provide a correlated prediction result here. In contrast, digzyme successfully predicted a weakly correlated result (r = 0.411, Figure 2, lower center). Even when limiting the data to the 46 amino acid sequences used in FoldX predictions, a weak correlation was observed (r = 0.273), which signifies a remarkable advancement in expanding the applicability of existing physical models. AntiBERTy and ESM-2 are protein language models, and the scores in Figure 2 represent the (pseudo) log-likelihood of the amino acid sequences [5]. This is an indicator of the sequence's probability or plausiblity, and some correlation with thermal stability was expected. However, the actual correlation was weak (correlation coefficients of 0.168 and 0.338, respectively). Therefore, as the authors suggest, if these language models are used for thermal stability prediction, it would be better to build a model specialized for the task through further training [5]. On the other hand, the low correlation of DeepSTABp, a dedicated regression model for predicting T_m (r =0.267, Figure 2, upper right), requires careful consideration. According to the original paper [6], the Pearson correlation coefficient for DeepSTABp's test data was 0.90, yielding significant results. Despite this, the low correlation with the NanoMelt dataset suggests that this model may have overfitted and has limitations in generalization ability (i.e., it is powerless against "unfamiliar data"). In contrast, the physical model adopted by digzyme is based on general physical principles, which have no inherent limits to the range of amino acid sequences they can be applied to. In fact, since digzyme's model gives a higher correlation than DeepSTABp, it can be said that the superiority of the physical model is demonstrated to some extent.

Now, the NanoMelt model, which shows a strong correlation (r = 0.862, Figure 2, lower right), raises concerns about potential overfitting. Unfortunately, since NanoMelt is a nanobody-specific model, it is difficult to re-evaluate its performance on other datasets containing arbitrary protein sequences. Therefore, the authors selected six new sequences from the camel nanobody sequence database and experimentally measured T_m to re-evaluate the model. The criteria for selecting the sequences were as follows:

　1.The dissimilarity with the most similar sequence in the NanoMelt dataset is at least 30%.
　2.The AbNatiV VHH-nativeness score (a measure of sequence likelihood) is 0.85 or higher.
　3.The predicted T_m in NanoMelt is either low (T_m < 61°C) or high (T_m > 73°C).

Of these six sequences, the three predicted to have low T_m did not express, but the three predicted to have high T_m did express, and successful T_m measurements were made. The error in the predicted results was approximately 1°C, demonstrating the high predictive performance of NanoMelt [5]. While this result does not definitively determine the presence of overfitting, it provides a clue to understanding the applicable range of the NanoMelt model.

Conclusion

In the context of enzyme discovery and modification, calculating the thermal stability of enzymes and creating an accurate ranking is crucial to reducing subsequent experimental processes. At digzyme, we tackle this challenge by applying a general-purpose physical model. In this blog, we evaluated our developed model by calculating prediction scores for two datasets. As shown in Case 1, we demonstrated that our model achieves practical accuracy in predicting the thermal stability of variants. However, as shown in Case 2, while we showed the superiority of our model over existing physical and machine learning models in comparing arbitrary amino acid sequences, the correlation with experimental values was weak, indicating room for further improvement.

Recently, there have been increasing successful cases of enzyme design using AI, with expectations for realizing novel folds (structures) and functions that were overlooked during the evolutionary process. However, the novelty of these designs makes predictions of physical properties, such as thermal stability, by machine learning challenging. Therefore, the importance of general-purpose physical models, such as those introduced in this blog, is expected to grow even more. At digzyme, we plan to continue actively advancing research and development to build more accurate and reliable models.

References

[1] CAGI5 Frataxin [Link]
[2] Savojardo et al. Hum. Mutat., 2019, 40(9), 1392 [Link]
[3] Lindorff-Larsen et al. Science, 2011, 334, 517 [Link]
[4] Berliner et al. PLoS ONE, 2014, 9(9), e107353 [Link]
[5] Ramon et al. bioRxiv, 2024 [Link]
[6] Jung et al. Int. J. Mol. Sci., 2023, 24(8), 7444 [Link]