
SyntheticGestalt (Shinjuku, Tokyo; CEO Koki Shimada; hereinafter "our company") is pleased to announce its selection for GENIAC (Generative AI Accelerator Challenge) *1, a project conducted by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO) aimed at strengthening Japan's generative AI development capabilities. Specifically, our company has been selected for the survey project on data utilization and generative AI application to accelerate generative AI development.
This survey project will simultaneously achieve the construction of a database 10 times larger than the current world's largest—encompassing 100,000 to 200,000 entries—and AI-driven computational acceleration of 30,000 to 100,000 times (reducing processing time from hours-to-days to seconds). This will dramatically accelerate computer simulation-based development in pharmaceuticals, agrochemicals, and chemical products. Eight leading Japanese companies representing the pharmaceutical, agrochemical, and chemical industries are collaborating on this initiative, aiming to strengthen the international competitiveness of Japan's chemical industry through a Japan-originated, world-class data infrastructure.
Our CEO, Koki Shimada, commented: "Just as ChatGPT transformed the world with AI for language, we are revolutionizing drug discovery and the chemical industry with 'AI for molecules.' Our company possesses technology recognized by global tech giants and a robust collaborative framework with leading companies in the pharmaceutical, agrochemical, and chemical sectors. Through this project, we will merge Japan's strength in chemistry with AI technology to build a data infrastructure that leads the world. This is not merely a technological development—it is a historic step toward reclaiming Japan's industrial competitiveness."
Project Overview

Selection Results Announcement Page: https://www.nedo.go.jp/koubo/CD3_100408.html
About SyntheticGestalt
SyntheticGestalt is a molecule-specialized AI technology development company founded in 2018. With the mission of "Develop an autonomous system to make valuable discoveries en masse" our company provides technology that dramatically improves R&D efficiency in the drug discovery and chemical industries. Our company has achieved world #1 rankings in all three critical drug discovery metrics—toxicity prediction, permeability prediction, and stability prediction—in global standard drug discovery benchmarks. Our company has been featured as an invited speaker at NVIDIA GTC 2024/2025 and recognized as a technology innovation company on Google's official blog, earning high praise from global tech companies. Our company has developed the world's largest molecular AI model, "SG4D10B" *5, trained on 10 billion data points. Our company has achieved 90% reductions in both development time and costs at multiple major corporations. Our company has been selected for two consecutive terms in GENIAC's foundation model development support program, promoted by METI and NEDO.
[Notes]
*1 GENIAC (Generative AI Accelerator Challenge)
The common name for a "project aimed at strengthening Japan's generative AI development capabilities" conducted by METI and NEDO. To accelerate the development and social implementation of generative AI in Japan, the project supports foundation model development, data infrastructure development for promoting generative AI utilization, and operates a community centered on participating organizations.
*2 FEP calculations (Free Energy Perturbation)
A state-of-the-art computational chemistry technique that predicts the binding strength (binding affinity) between molecules and target proteins with high precision through physics-based simulation. While FEP enables prediction of compound efficacy prior to experimentation in the drug discovery process—promising significant efficiency gains—the computational cost of approximately 24-72 GPU hours per compound has been a barrier to practical implementation. This project will accelerate these calculations by 30,000 to 100,000 times to just seconds using AI technology.
*3 Target proteins
Proteins that are the targets of drug action. Therapeutic effects are achieved by inhibiting or activating proteins involved in diseases. In drug discovery, it is crucial to search for compounds that bind strongly to target proteins while not binding to other proteins. In agrochemical development, proteins involved in pest or weed growth serve as targets.
*4 Binding affinity
A quantitative indicator of "how strongly a molecule interacts with a target protein." In drug and agrochemical development, binding to target proteins is a prerequisite for efficacy, and it is also key to functional expression in cosmetics and materials fields, making it the most fundamental and important metric.
*5 SG4D10B
The world's largest molecule-specialized AI foundation model developed by our company. Trained on 10 billion compound data points, it can make predictions considering the complex three-dimensional structures of molecules. "4D" refers to our company's proprietary molecular representation technology that considers multiple conformations (3D + time axis) per compound.
