Simulation of the Direct Cell Conversion Process: AI Predicts Small Molecule Compounds that Induce Cell Reprogramming

A research group led by Associate Professor Momoko Hamano from the Graduate School of Computer Science and Systems Engineering at the Kyushu Institute of Technology, in collaboration with Professor Yoshihiro Yamanishi from the Graduate School of Informatics at Nagoya University (Tokai National Higher Education and Research System), has developed a novel method to predict the optimal combination of small molecule compounds*3 that induce "direct reprogramming"*2—the direct conversion of cells into other cell types without going through iPS cells*1.

Key Points
- Developed an AI technology that predicts small molecule compounds capable of inducing direct reprogramming, converting somatic cells directly into different types of cells.
- Succeeded in simulating the cell conversion process, classifying it into multiple stages such as early, middle, and late, and predicting the appropriate combination of small molecule compounds for each stage.
- The proposed method is expected to lead to improvements in the efficiency and safety of cell production for cell therapies, including the field of regenerative medicine.

This research group has developed an AI technology that predicts small molecule compounds (such as drugs and chemicals) that induce direct reprogramming to directly convert somatic cells into other cell types. In direct reprogramming, there is a strong demand for the establishment of methods that induce cell conversion using highly safe small molecule compounds. Therefore, with this proposed method, the team developed an AI technology that can easily predict the small molecule compounds required for each stage of cell conversion. First, the induction process of direct reprogramming was reproduced through simulation using single-cell level gene expression data. Next, the dynamic molecular mechanisms of cell conversion were clarified by dividing the cell conversion process into early, middle, and late stages, and examining the gene expression patterns that change at each stage. Finally, the team succeeded in predicting the combinations of small molecule compounds that control the molecular mechanisms changing at each stage using an optimization algorithm. The proposed method makes it easy to predict small molecule compounds that induce direct reprogramming in a step-by-step manner, and is expected to lead to improved efficiency and safety of cell production in the future field of regenerative medicine.

These research findings will be published in Communications Chemistry on May 18, 2026.

■ Background and History of the Research
Direct reprogramming is a technology that directly converts somatic cells into other types of cells without going through an undifferentiated cell stage like iPS cells. Because cells can be produced at low cost in a short period, it is anticipated to be an innovative technology for future regenerative medicine. However, because conventional methods are induced by introducing genes encoding transcription factors into the original cells, there has been a risk of tumorigenesis associated with gene transfer. To avoid the risk of cancer, direct reprogramming methods using small molecule compounds (drugs, chemicals, etc.) have been gaining attention (Figure 1).

Experimentally identifying the optimal combinations of small molecule compounds that induce direct reprogramming involves enormous experimental costs and is extremely difficult. Additionally, development of induction methods using small molecule compounds has lagged behind because their cell conversion efficiency is lower compared to the introduction of transcription factors, resulting in the conventional focus on methods utilizing transcription factors. Therefore, there is an urgent need for technologies that effectively utilize the recently accumulating big data in life sciences to capture the induction process of cell conversion at the single-cell level and identify small molecule compounds that control the mechanisms of cell conversion.

Figure 1: Direct reprogramming that directly converts cells into other types of cells using small molecule compounds (Part of the illustration is cited from TogoTV)

■ Content of the Research
The research group consisting of Associate Professor Momoko Hamano, graduate student Ryofu Ito (at the time), graduate student Ryota Kawasaki (at the time), graduate student Hikaru Watanabe, and graduate student Arisa Matsuo from the Graduate School of Computer Science and Systems Engineering at the Kyushu Institute of Technology, in joint research with Professor Yoshihiro Yamanishi from the Graduate School of Informatics at Nagoya University, has developed an AI model that predicts small molecule compounds that induce direct reprogramming.

In the proposed method of this study, the induction process of direct reprogramming was first reproduced by simulation. Then, by dividing the cell conversion process into stages such as early, middle, and late, and examining the gene expression patterns that change at each stage, the dynamic molecular mechanisms of cell conversion were clarified. Finally, the combination of small molecule compounds that control the molecular mechanisms of cell conversion changing stage by stage was predicted using an optimization algorithm (Figure 2).

Figure 2: The proposed method predicting the combinations of small molecule compounds that induce direct reprogramming stage by stage
The process of inducing direct reprogramming was simulated. The cell conversion process was divided into stages such as early, middle, and late, and by examining the gene expression patterns that change at each stage, the dynamic molecular mechanisms of cell conversion were clarified.