[By ktsdesign/Adobe Stock]
Google has made significant investments in AI and machine learning for life sciences. One recent example is Med-Gemini, a family of advanced AI models specifically designed for medical applications and launched earlier this year. These models are multimodal, capable of processing and interpreting diverse data types, including text, images, audio, and video, within medical contexts. Google Cloud also offers specialized services like Healthcare Data Engine and Vertex AI Search for Healthcare, enabling life sciences organizations to manage and analyze complex healthcare data.
Google is also forging a growing number of partnerships in areas ranging from dynamic protein modeling to synthetic data generation.
Superluminal is rethinking drug discovery with dynamic protein modeling
Shweta Maniar
One Google partner, Superluminal Medicines, is a generative biology and chemistry company accelerating drug discovery through AI-powered dynamic protein modeling. Recognizing that proteins are not static but constantly in motion, Superluminal has developed a platform that captures this dynamic behavior. The company launched in August 2023 with a $33 million seed round. It won the backing of prominent firms like RA Capital, Insight Partners, NVIDIA, and Gaingels.
Superluminal’s platform creates candidate-ready compounds with unprecedented speed using a combination of deep biology and chemistry expertise, machine learning, and proprietary big data infrastructure. The predict-design-test architecture accurately models protein shapes and designs highly selective compounds to target precise structural changes for therapeutic effect.
Superluminal leverages Google Cloud’s computing power to analyze multiple protein structures and integrate them into its dynamic protein models. “I often use the analogy of combining a bunch of individual photos to create a movie — a continuous, moving image,” Maniar said. “These dynamic models provide a much more accurate picture of how proteins function within the cell or body, enabling the design of more precise drug interventions.”
Superluminal’s platform focuses on creating candidate-ready compounds that target G protein-coupled receptors (GPCRs) with small molecule therapeutics. GPCRs are a large family of cell surface receptors crucial in various physiological processes and are important drug targets.
Using advanced computational methods and Google Cloud’s compute, Superluminal models GPCRs more accurately by creating dynamic protein models that capture their constant motion. “Superluminal went beyond sequence to structure, generating multiple different conformations or shapes,” Maniar said. “They then enhance these publicly available foundation models based on their expertise and the applications they need from Google.”
Ginkgo Bioworks is building foundational AI models for biotech
Ginkgo Bioworks is also partnering with Google Cloud to leverage large language models (LLMs) for a variety of biotech applications. Last year, Ginkgo and Google Cloud announced a partnership to build LLMs using the Vertex AI platform and Ginkgo’s biological data (over 2 billion unique protein sequences and vast functional assay data). An article from last year noted the goal of developing an LLM model that can “speak DNA.” This collaboration focuses on diverse biotech applications, including genomics, protein function, and cell engineering, with an initial focus on creating a foundational model for proteins.
“Ginkgo Bioworks is leveraging Google Cloud to fine-tune and train large language models for a variety of tasks,” Maniar explained. “They are using AI to build representations of proteins and DNA across multiple organisms and cell states to create broad tools for various applications.” Ginkgo’s goal is to use generative AI to create comprehensive representations and “empower the biotech system,” she said. “Rather than developing drugs themselves, Ginkgo Bioworks is focusing on creating models that are accessible to partners through APIs, software packages, and collaborative partnerships.”
Bayer’s work with synthetic data
Beyond its work with Superluminal Medicines and Ginkgo, Google Cloud is also actively collaborating with Bayer across multiple life science areas. One key area of focus is the development of synthetic images for oncology using histological images. Bayer is working on synthetic images for oncology, created from histological images. This is important because there’s limited data available to train algorithms, particularly in the rare disease space. These synthetic images have been extremely useful for Bayer in both oncology and radiology,” explained Maniar.
Bayer is also tapping generative AI to develop an AI-powered radiology platform that assists radiologists in identifying anomalies, accessing patient histories, and making more informed diagnoses. Another focus area is the substantial automation of regulatory processes. “Bayer has also enabled the company to automate the population of 70 to 80% of their regulatory dossiers, streamlining their entire regulatory process,” Maniar said.
Google Cloud’s AI ambitions just warming up
Google Cloud’s foray into AI-powered drug discovery has shown early promise, including faster identification of promising drug targets and streamlined clinical trials. “It’s just the beginning,” Maniar emphasizes. “We’re all experimenting. We’re looking at Google’s capabilities in AI, and then we’re looking at the industry specifically, and thinking very intently about what can we do for this industry. How do we make our technology most useful?”
In the long run, Maniar envisions a future where generative AI transforms drug development. “We’re really making, taking a stance that generative AI really does have the potential to significantly reduce time and cost, to bring new therapies to market faster at, you know, to the right patients at the right time, with the speed that organizations are always looking for,” she explained. This includes the potential to unlock new drug classes and treatment options previously unimaginable.
This journey requires substantial collaboration. “It’s been a very, very big learning process, both for our partners as well as for, you know, us, as we’re continuing to understand the use cases and the applications of the technology of AI in this, in the industry,” said Maniar. “Which is why we always say, right, it’s not, you know, we’re not looking at Ginkgo or Bayer or any of these others as customers. We’re looking at them as partners, because we are learning from each other.”
Filed Under: Drug Discovery, machine learning and AI
