AI
Using our drug discovery platform, ibVIS®, which integrates informatics with experimental biology, we have advanced our research into small molecule and nucleic acid drugs targeting mRNA. As our drug discovery initiatives progressed, data accumulated and AI technologies matured, enabling ibVIS® to evolve into aibVIS—an AI-integrated platform that combines AI with experimental biology.
The hallmark of aibVIS is its use of numerous “specialized AIs” tailored to specific objectives, orchestrated together.
Drug discovery requires different areas of expertise at each stage of the process. To address the unique needs and challenges of each step, we chose to combine multiple specialized AIs, each strictly focused on a specific goal. By intentionally limiting their scope, these specialized AIs achieve high performance within their domain, remain compact, and are cost-effective to build. This practical, realistic approach differs from the strategy of major IT companies aiming to create a single, general-purpose AI, and is particularly well-suited to companies like ours. Thanks to our proprietary drug discovery data and a close, collaborative setup where experimental scientists and software developers work side by side, we can rapidly create specialized AIs aligned with frontline needs.
In other words, aibVIS represents a new form of drug discovery platform in which diverse specialized AIs each exercise their expertise and integrate organically with our researchers.
The drug discovery platform, aibVIS, incorporating multiple specialized AIs
Technically, we utilize “rule-based AIs” together with “data-driven AIs.”
Rule-based AIs encode scientific theories as well as expert knowledge into explicit rules, making decisions accordingly. They are particularly useful in emerging areas with limited data available. Since our founding, we have pursued mRNA-targeted drug discovery by applying rule-based AIs to mRNA targets whose data was scarce.
By contrast, data-driven AIs, which are the current mainstream, train machine learning and deep learning models on large datasets. As our internal research and collaborative research with pharmaceutical partners have accumulated ample data on mRNA-targeted drug discovery, we began implementing data-driven AIs tailored to specific drug discovery tasks.
One of our data-driven AIs is already enhancing efficiency in the screening stage of mRNA-targeted small molecule drug discovery.
In our machine learning–assisted screening approach called AISLAR (AI-augmented Iterative Screening Libraries Against RNA targets), we statistically analyze results from the first batch (the initial set of compounds evaluated), and based on those findings, the AI selects the compounds to be assessed in the next batch. By iterating this cycle, “perform statistical analysis, then select compounds via AI,” we found that screening only about 20% of the entire library yields more than half of the hits that would be obtained by exhaustively screening the entire library. In other words, AI can boost screening efficiency by roughly twofold or more.
Screening workflow and example performance evaluation with AISLAR
Thus, building upon and refining our specialized rule-based AIs cultivated since our founding and progressively introducing specialized data-driven AIs as our business advanced, we have developed our drug discovery platform into aibVIS. Going forward, while preserving our existing advantages, we will further improve efficiency to drive both our platform and pipeline businesses by ambitiously developing AI technologies specialized for drug discovery challenges and embedding them into our drug discovery processes.
We believe this approach is suitable for a small biotechnology company like ours and commensurate with its scale. We have a system in place that enables the agile development of specialized AIs with our experimental scientists and software developers together in close proximity—often being the same individual, resulting in a “co-creative” drug discovery platform in which researchers and AIs evolve together. Our model, where each specialized AI demonstrates its expertise while harmonizing with human collaborators, naturally resonates with the Japanese cultural concept of “Yaoyorozu no Kami”—the belief that the world is shaped by myriad deities, each fulfilling their own specific role. We are proud to be a biotechnology company originating from Japan, and will continue to leverage these strengths as we grow.
