Although over the past ten years Google DeepMind has been the leading company in applying AI to scientific research, it now seems to have fallen behind, while Anthropic is experiencing its golden moment and has decided to commit even more to drug discovery with the help of AI.
The company led by the Amodei brothers has in fact introduced Claude Science, positioning itself to inherit DeepMind’s legacy in this sector. It is not just a matter of scientific ambition: Dario Amodei, like Demis Hassabis of Google DeepMind, has a research background (unlike Sam Altman, who is more entrepreneurially oriented). Moreover, among the new arrivals at Anthropic is John Jumper, a researcher who shared the 2024 Nobel Prize in Chemistry with Hassabis.
A DIGITAL LABORATORY FOR SCIENTIFIC RESEARCH
Claude Science is the new product from Anthropic designed for scientific research, particularly in biology, chemistry, and pharmaceutical development. It is not a new artificial intelligence model, but a work environment that integrates existing tools into a single platform. The goal is to reduce the fragmentation typical of scientific work, which often requires switching between different databases, analysis software, and computational tools.
The system allows scientists to ask questions in natural language and receive answers without having to consult many sources separately. Claude Science combines over 60 scientific databases with programming and analysis tools, enabling the management of complex workflows in a single digital space. Among the main features is also the ability to generate and analyze scientific content such as three-dimensional protein structures, chemical representations, and genomic data.
HOW IT WORKS IN PRACTICE
The operation of Claude Science, explains TechCrunch, is based on a central assistant that coordinates research activities like a project manager. This system can delegate tasks to specialized sub-agents and interact with tools already configured for specific disciplines such as genomics, structural biology, and chemistry.
A significant part of the system is dedicated to traceability of results. Every output, such as a graph or a molecular structure, is accompanied by the code that generated it, the execution environment, and the history of operations. In this way, scientists can verify the path that led to a particular result and modify it simply by describing the desired corrections in natural language.
Anthropic has also emphasized that the platform is designed to improve research reproducibility and reduce errors related to the use of unverifiable sources, a growing problem in AI-assisted scientific work.
PHARMACEUTICAL RESEARCH AND INDUSTRIAL USES
Claude Science, writes the MIT Technology Review, was mainly designed for the life sciences and drug development sector. According to Anthropic, the system is already used to identify therapeutic candidates and analyze rare and neglected genetic diseases, with direct applications in molecular and cellular biology.
During a presentation, Alexander Tarashansky demonstrated how the system was able to identify possible drug candidates for phenylketonuria, a rare genetic disease. Anthropic also announced that it will internally use the system to start preclinical drug discovery programs, focusing on areas that the traditional pharmaceutical market tends to overlook.
“These are areas that are outside the spectrum of what the traditional pharmaceutical sector considers attractive targets, but still have a real impact on health,” explained Eric Kauderer-Abrams, head of life sciences at Anthropic.
COMPETITORS’ STRATEGIES
The launch of Claude Science takes place in a competitive context where various companies are adopting different strategies to apply AI to scientific research.
OpenAI has developed a more selective approach with models like GPT-Rosalind, designed for biological research and drug discovery, but distributed only to qualified enterprise clients and subject to security checks and limited access. Initial partners include Amgen, Allen Institute, Moderna, Thermo Fisher, and Novo Nordisk.
Google DeepMind, on the other hand, maintains a different position, based on proprietary scientific models like AlphaFold and AlphaGenome, used as fundamental tools for computational biology. Its Gemini for Science platform integrates these models with over 30 scientific databases, creating a closed and highly specialized ecosystem.
Anthropic has chosen a more open strategy, making Claude Science available in beta to all paying users, from Pro plans to Enterprise. At the same time, it has started collaborations with pharmaceutical companies like Novo Nordisk and research institutes like the Allen Institute, which have already used Claude for activities such as drug discovery and scientific literature synthesis.
A MARKET BETWEEN SCIENCE AND INDUSTRY
The pharmaceutical sector has become one of the main fields of application for artificial intelligence. Companies like Eli Lilly have invested in computing infrastructure and collaborations with specialized firms like Insilico Medicine to accelerate the search for new drugs. In this context, Claude Science fits as a tool designed to reduce the time of the initial phase of molecular discovery and scientific analysis.
It is no coincidence that the presentation of Claude Science took place in San Francisco before an audience composed of pharmaceutical companies, biotech startups, and research centers. On stage, alongside Anthropic CEO Dario Amodei, was Vas Narasimhan, CEO of Novartis and a member of Anthropic’s board of directors, while an intervention by Chris Boerner, CEO of Bristol Myers Squibb, was also planned.
The announcement, reports Bloomberg, had an immediate impact on financial markets as well. Companies specialized in drug discovery experienced strong stock market fluctuations: Schrödinger lost up to 8.3%, Recursion Pharmaceuticals fell by 3.3% before recovering part of the losses, while IQVIA Holdings dropped as much as 2.3%, subsequently erasing the decline. This demonstrates that investors are attentive to the growing role of artificial intelligence in pharmaceutical research.
