Ginkgo Bioworks Holdings, Inc., a leader in cell programming and biosecurity, has unveiled two innovative tools designed to streamline drug development for pharmaceutical and biotech companies. Building on last year’s partnership with Google Cloud, these offerings aim to significantly enhance the way companies develop new medicines.

The first is a groundbreaking protein large language model (LLM) developed in collaboration with Google Cloud Consulting. This LLM will give both individual researchers and enterprise companies access to insights derived from Ginkgo’s proprietary data, empowering them to advance their drug discovery efforts. The second offering is Ginkgo’s new model API, which brings sophisticated biological AI models directly to machine learning scientists. The API is now publicly available on Ginkgo’s website, while the protein-based LLM will soon be accessible via Google Cloud’s Vertex AI Model Garden.

A Major Step for AI in Drug Discovery

Jason Kelly, CEO of Ginkgo Bioworks, expressed enthusiasm about the potential of these new tools: "I’m excited to see how the community builds on top of these models and our API. AA-0 is the first model we’ve released that is trained on Ginkgo’s proprietary data. We’re opening it up to data scientists and bioinformaticians to build new models and applications." Kelly also highlighted the cost-effectiveness of Ginkgo’s offerings, with customer-friendly terms like low token prices, no royalties, and a firm commitment not to reuse customer data.

These offerings mark a significant step forward in how Ginkgo is enabling the life sciences industry to improve and accelerate the drug development process.

The Protein LLM: A Powerful Tool for Biotech

The new protein LLM is built on Google Cloud’s Vertex AI, leveraging Ginkgo’s vast proprietary dataset. This tool, and others like it, will allow researchers and companies to harness the power of AI to analyze and understand complex protein structures. By doing so, they can optimize lead identification and accelerate the discovery of new therapeutics. With access to Ginkgo’s private data, companies will be able to unlock hidden patterns and uncover potential therapeutic targets that were previously out of reach.

The second launch, Ginkgo’s open API, is designed for ease of use and scalability, providing researchers with access to advanced AI models trained on protein and DNA data. Its first release, ginkgo-AA-0-650m, is a machine learning model trained on over two billion proprietary Ginkgo protein sequences. With this API, scientists can easily access tools that were previously only available internally, opening new possibilities for drug discovery and synthetic biology.

Chris Sakalosky, Vice President of Strategic Industries at Google Cloud, praised the development: “Ginkgo’s new protein LLM and open API mark a major step forward in making advanced AI tools accessible for drug discovery and biological research.”

Looking Ahead: Expanding AI Capabilities

Ginkgo is already working on an array of models, spanning machine learning techniques like language modeling and diffusion for conditional design. The company’s first protein LLM release will support two key use cases: sequence generation via masked language modeling, and embedding calculation to extract valuable representations for downstream tasks.

Over the coming year, Ginkgo plans to roll out more models and expand its API’s capabilities, creating a comprehensive suite of tools that will address complex challenges in drug discovery, synthetic biology, genomics, and beyond.

Ankit Gupta, General Manager of Ginkgo AI, emphasized the company’s commitment to accessibility: “Flexibility is everything. Alongside our first proprietary model, you’ll also have access to publicly available models like ESM2. Our API comes with competitive pricing and a free tier to make it easy for researchers to experiment and get predictions without high fees.”

With its robust tools and customer-friendly pricing, Ginkgo is positioning itself at the forefront of the generative biology revolution, enabling scientists to design thousands—or even millions—of sequences at a time.