AgentAI's AI Drug Discovery Platform 'MolForge' Selected for NVIDIA Inception

MolForge, an artificial intelligence (AI)-based drug candidate discovery platform developed by Digital Biology, a subsidiary of AgentAI, has been recognized for its technological competitiveness after being selected for NVIDIA’s global startup support program. The company plans to accelerate the advancement of its platform by leveraging AI computing infrastructure and to expand collaborations with global pharmaceutical and biotech companies.

On July 1, AgentAI announced that MolForge, Digital Biology’s AI small molecule drug discovery platform, had been selected for NVIDIA’s startup accelerator program, NVIDIA Inception.

This marks the second time a subsidiary of AgentAI has joined the NVIDIA Inception program. Previously, AgentAI Labs, which develops the AI robotics safety verification platform RoboGate, was also selected for the same program.

With this selection, Digital Biology will participate in NVIDIA’s global AI ecosystem and aims to strengthen its research and development capabilities and expand its business based on its AI drug development technology.

A company official commented, “It is significant that our self-developed AI platforms have been consecutively recognized for their technological prowess in future growth industries such as robotics and drug development.”

NVIDIA Inception is a program that supports innovative companies in advanced technology fields such as AI, biotech, and robotics, providing GPU computing infrastructure, technical support, and global networking. Digital Biology plans to use this opportunity to enhance its AI research environment and expand joint research and business collaborations with global pharmaceutical and biotech companies.

Drug development typically requires more than 10 years and massive investment, and only a limited number of candidates from a vast pool actually become therapeutics. As a result, the field of ‘digital biology,’ which uses AI and high-performance computing to innovate the drug development process, is emerging as a new growth area.

MolForge is an AI-based platform that supports the entire process of drug candidate discovery, from small molecule compound design and validation to prioritization. Its key feature is that it is not limited to a specific drug but has a platform structure that allows expansion across various disease areas.

MolForge designs new small molecule compounds by training on approximately 1.6 million drug data points using its proprietary generative AI model, MolForge-Gen. Of the generated molecules, 97.5% exhibited chemically valid structures, and a multi-verification system using more than seven independent AI models—including those for molecule generation, binding prediction, toxicity, drug-likeness, and synthetic feasibility—was applied.

To date, MolForge-Gen has generated 2,190,168 candidate molecules. Of these, 71,651 candidates have been selected after evaluating drug-likeness (ADMET), novelty, and synthetic feasibility, and the company operates an AI funnel system that narrows these down to 2,401 substances for priority review.

This system allows AI to automatically review large-scale compound data that would be difficult for humans to analyze directly and efficiently identify promising drug candidates.

Digital Biology has also implemented verification technologies to increase the reliability of AI predictions. The conformal prediction technology, which indicates prediction confidence, achieved a hit rate of 86.5% in external evaluations, and 99.5% of the protein-molecule binding structures predicted by AI passed the physicochemical validity assessment of PoseBusters, an internationally recognized validation tool.

Additionally, the company ensures transparency and reproducibility in its research processes by pre-registering AI prediction results, along with time stamps, in public repositories before experiments are conducted.

More than 95% of the candidate molecules generated by MolForge consist of novel structures distinct from existing drugs, thus securing patent competitiveness. The company also continues to verify AI performance using global open benchmarks such as TDC, Polaris, and PLINDER.

Currently, Digital Biology is simultaneously conducting five small molecule drug discovery projects: TYK2, TNIK, CDK4, CDK6, and EGFR. An automated, 24-hour AI research and development system covering the entire process from candidate generation to evaluation and retraining has been established to improve drug development efficiency.

A Digital Biology official stated, “This selection for NVIDIA Inception demonstrates that our AI drug discovery platform’s competitiveness is recognized within the global AI ecosystem. It is further validation of the AI platform technology held by the AgentAI group.”

They added, “MolForge is an AI platform that can be expanded to various diseases beyond single-drug development. We will grow into a global AI drug discovery platform company by expanding joint research, technology transfers, and strategic partnerships with global pharmaceutical companies.”

AgentAI will leverage its entry into the NVIDIA AI ecosystem to advance the technology of the MolForge platform, expand global collaborations, and strengthen its competitiveness in the AI-driven drug development market.

This content was produced with the assistance of AI translation services.