日本語 
Bringing a new medicine to patients can take over a decade. The bill to get it there can also climb past $1 billion. Each time a trial fails, it is not just a scientific issue. It also hits budgets and plans. That matters a lot for countries that want to keep a strong role in drug research across the world.
Japan is now trying to meet this challenge with artificial intelligence. The goal is to work on the problem earlier and more directly. AI is already changing how teams look for drug targets. It also affects how researchers shape candidate molecules and how they test large sets of compounds. It can even help estimate how a candidate may act in the body.
Another change is just as important. Japanese pharma is moving away from one-off projects. Instead, more groups want shared systems, bigger computing resources, and closer ties with partners. This piece reviews why Japan is leaning on AI, and how machine learning is altering each stage of the drug discovery workflow. It also points to what firms like Chugai have been doing. The takeaway is that the next edge may go to groups that know how to work with AI day to day, not just those that add it as a tool.
The Challenges Driving AI Adoption in Japan
Drug discovery has a productivity problem that technology alone cannot hide. The industry has spent years trying to develop medicines faster, yet the process remains uncertain, expensive and heavily dependent on experiments that can fail late. This tension is often described through Eroom’s Law, the observation that drug research productivity has struggled even as scientific knowledge and technology have advanced.
Japan faces another layer of pressure. Its pharmaceutical industry is not operating in a vacuum. JPMA says Japan is experiencing a structurally declining share of global biopharmaceutical R&D and investment, while early-stage pipelines and clinical-trial activity are increasingly shifting elsewhere. That creates a strategic problem. If research activity keeps moving outward, Japanese pharma needs more than good science. It needs a better way to turn scientific knowledge into viable medicines.
The country’s demographic reality makes that pressure harder to ignore. An ageing population means healthcare demand is changing, while difficult disease areas such as oncology and immunology require increasingly complex research. Researchers therefore need to examine more possibilities without allowing the discovery process to become slower and more expensive.
The government is responding as well. Japan’s FY2026 budget allocates ¥44.4 billion to promoting innovation in pharmaceuticals and medical devices, including ¥7.6 billion for strengthening the drug-discovery foundation support project. The message is fairly clear. Drug discovery is no longer being treated only as a pharmaceutical-company problem. It is becoming a national innovation priority.
That changes the role of AI. The objective is not to make scientists work harder. It is to help them eliminate weaker possibilities earlier, focus experiments more intelligently and compress parts of the discovery cycle that have traditionally consumed enormous amounts of time.
The Core Technologies Powering the Revolution
AI drug discovery becomes easier to understand when the technology is connected to the actual work researchers perform. Instead of treating AI as one giant tool, Japanese research programmes are applying different models to different bottlenecks.
Generative AI and Molecular Design
Traditional molecular design involves researchers creating, modifying and testing potential compounds through repeated cycles. Generative AI can change that starting point. Algorithms can suggest new molecular structures based on desired properties, giving researchers a larger set of candidates to evaluate.
The value is not simply generating more molecules. More candidates can create another problem if researchers cannot separate promising ones from poor options. Therefore, the real advantage comes when generation is connected to prediction and experimental validation. AI can help narrow the search before researchers commit time and resources to physical testing.
High-Speed Screening and Supercomputing
Computing power is becoming just as important as the model itself. Large-scale virtual screening and molecular docking allow researchers to evaluate potential interactions computationally before moving into laboratory work.
Japan is building infrastructure to support that shift. An AI supercomputer with approximately 2,000 of the latest GPUs is planned for full-scale operation in FY2026, supporting research that includes drug discovery. That matters because advanced AI models and molecular simulations require serious computing capacity.
The bigger shift is from asking researchers to search a huge chemical space manually toward using machines to reduce that space first. The scientist still makes the important decision. AI simply helps make the search less blind.
Predictive Analytics for Pharmacokinetics and Toxicity
A promising molecule can still fail if it behaves badly inside the body. This is where predictive analytics becomes important.
Machine learning can help researchers predict pharmacokinetic behavior and identify potential toxicity earlier. Japan’s integrated drug-discovery AI platform provides a strong example. Approximately 15 million on/off-target, ADME and toxicity data points from 17 pharmaceutical companies were used in the platform, which was deployed to those companies.
That illustrates something important about AI drug discovery. The technology becomes far more useful when it can learn from the biological and pharmaceutical information that researchers already possess. The challenge then shifts from simply building better models to making useful data available in a secure and practical way.
How Japanese Pharma Is Applying AI
The strongest evidence of AI drug discovery in Japan comes from what pharmaceutical companies are doing with it, rather than what the technology promises to do.
Chugai Pharmaceutical
Chugai has been developing AI applications across its research activities, including candidate exploration, pharmacokinetic prediction, target identification, molecule design and process automation. Its work with Preferred Networks has also pushed machine learning and deep learning deeper into drug research.
One of its notable initiatives is MALEXA. The system uses machine learning on antibody amino-acid sequences to propose candidate sequences and support optimisation based on binding activity, pH dependency and physical properties.
That changes the nature of the research task. Instead of relying entirely on scientists to manually work through possible antibody sequences, machine learning can help identify combinations that deserve attention. Researchers can then test, interpret and refine those suggestions.
The distinction matters. AI is not becoming the medicinal chemist. It is becoming another layer of scientific capability.
Chugai’s approach also points toward a broader shift in pharmaceutical research. AI is not being placed in one isolated corner of the organization. It is gradually being connected to discovery, prediction, automation and decision-making. That is where the real productivity opportunity sits.
Why Open Innovation Is Becoming Essential
The idea that one pharmaceutical company can build every part of its AI drug discovery stack internally is becoming harder to defend.
Modern drug research needs biological expertise, quality data, AI models, computing infrastructure and laboratory validation. Those capabilities rarely sit under one roof. As a result, the competitive advantage increasingly comes from connecting them.
Japan’s model reflects that reality. Pharmaceutical companies can contribute domain knowledge and proprietary research data. Technology companies can bring computing and AI expertise. Universities can contribute new scientific methods. Research organizations can build infrastructure that individual companies may struggle to create alone.
This also explains why shared platforms matter. The AMED example is not simply about the size of its dataset. The involvement of 17 pharmaceutical companies shows how collaboration can create a broader base for AI drug discovery while allowing companies to participate in a common research ecosystem.
The same principle applies to computing. High-performance infrastructure becomes more valuable when researchers from different disciplines can use it to solve different parts of the same scientific problem.
For Japanese pharma, this may become one of the biggest advantages of the AI transition. The goal is not for every company to own the smartest model. The goal is to build an ecosystem where better models, better data and better experiments can reinforce each other.
The Future of the Medicinal Chemist and AI
The next phase of AI drug discovery will not be defined by whether machines replace scientists. That question is too simplistic.
The more useful question is where scientists should spend their time when machines can handle more of the search.
Also Read: Aisin and JEOL Develop Portable Refrigerant-Free NMR System
A medicinal chemist working with AI can test ideas faster, interrogate a wider set of possibilities and focus more attention on candidates that deserve experimental validation. Yet scientific judgment remains critical because an AI-generated candidate is only a hypothesis until evidence supports it.
Japan’s regulatory environment will also matter. As AI becomes more involved in drug design and development, regulators will need confidence in how these systems are used, how their outputs are evaluated and how human decisions remain accountable.
That is why AI drug discovery should not be viewed as another technology trend. Japan is facing pressure to make pharmaceutical research more competitive, while the tools available to researchers are becoming dramatically more capable. The companies that benefit most will not simply adopt AI. They will redesign the discovery process around it, while keeping scientific judgment firmly in human hands.
English