Galixir Leverages NVIDIA GPU Acceleration to Revolutionize AI-Driven Drug Discovery

• In this case, by leveraging NVIDIA GPUs, CUDA, and CuDNN, Star Pharma Technology has increased the efficiency of training models to generate drug candidate molecules with independent intellectual property rights and improved druggability from hundreds of millions of compounds by more than tenfold, while reducing the time required for predicting compound reaction pathways to the second level.

• This case primarily utilizes NVIDIA GPUs, CUDA, and cuDNN.

StarPharma Technologies is a startup leveraging AI technology to accelerate new drug development. By integrating multiple deep learning modules, the StarPharma AI platform significantly reduces the time and cost associated with new drug development while providing comprehensive patent protection for products under development. The company employs deep learning algorithms to build multiple preclinical drug development modules, substantially shortening the timeline required for drug discovery and development.

The discovery of new drugs is crucial to human health. Traditional drug development methods have only explored 10¹⁰level compound space. Drug R&D involves huge investments, long timelines, and low success rates; bringing a drug to market often requires over a decade of development and billions of dollars in investment. According to a Deloitte research report, the return on investment for global new drug development has fallen to a historic low of just 1.9%. There is an urgent need for new approaches to transform and accelerate the drug development process.

By integrating graph neural networks, reinforcement learning, and computational chemistry, the Star Drug team designs and screens drug candidate molecules. Leveraging NVIDIA GPUs, CUDA, and CuDNN, Star Drug Technology has developed multiple models to perform multimodal big data learning on known molecular structures. These models rapidly compute based on protein multi-level structures, compound group properties, and combination patterns, thereby generating drug candidate molecules with independent intellectual property rights and improved druggability from hundreds of millions of compounds. Compared with traditional CPU-based training, GPU acceleration enhances model training efficiency by more than tenfold.

In addition, Star Pharmatech leverages NVIDIA GPUs to process massive volumes of chemical reaction data, conducting further synthetic feasibility screening on molecules designed by AI models. This capability reduces compound reaction pathway prediction to the second level, creating more possibilities for humanity to explore a broader universe of compounds.

The traditional drug development and screening process takes years, but with the acceleration provided by NVIDIA GPUs, AI-powered drug screening can be completed in just a few days.

“The application of AI in drug R&D hinges on breakthroughs in algorithms and computing power. We are delighted that in this era, alongside the rapid advancement of deep learning algorithms, NVIDIA’s powerful GPU computing platform empowers us to explore more complex and effective models, thereby facilitating the practical implementation of AI in drug discovery,” said Dr. Li Chengtao, Founder and CEO of Xingyao Technology.。