Skip to content

Google DeepMind Science Skills

Source: Google DeepMind Science Skills

TL;DR

Google DeepMind has released an open-source collection of agent skills for scientific research tasks spanning genomics, structural biology, cheminformatics, and literature search. The toolkit integrates with AlphaGenome, AlphaFold DB, UniProt, and over 30 other scientific databases, and is designed for the Google Antigravity platform. It includes structured instructions, scripts, and resources for specialized scientific tasks. The project requires the uv Python package manager, some skills need API keys (e.g., AlphaGenome, OpenAlex), and it is licensed under Apache 2.0 — though it is not an official Google product.

What Are Science Skills?

The DeepMind Science Skills repository provides a set of modular, reusable agent capabilities designed to automate and assist with specialized scientific workflows. Each "skill" encapsulates domain knowledge about a particular scientific task — how to query specific databases, how to interpret results, what parameters to use, and how to chain multiple operations together.

The skills are designed to run on Google Antigravity, DeepMind's agent platform, but the code and instructions are open-source and could potentially be adapted to other agent frameworks.

Covered Domains

The skills span multiple scientific domains:

  • Genomics: Skills for working with genetic sequences, genome annotations, and variant analysis, integrating with AlphaGenome
  • Structural Biology: Skills for querying AlphaFold DB, analyzing protein structures, and predicting protein folding outcomes
  • Cheminformatics: Tools for working with molecular structures, chemical databases, and drug discovery pipelines
  • Literature Search: Skills for finding and synthesizing relevant scientific papers, integrated with OpenAlex and other bibliographic databases

Database Integrations

The project connects to 30+ scientific databases, including:

  • AlphaGenome — DeepMind's genomic analysis platform
  • AlphaFold DB — Repository of protein structure predictions
  • UniProt — Universal Protein Resource
  • OpenAlex — Open catalog of scholarly papers
  • PDB (Protein Data Bank)
  • Various NCBI databases
  • Specialized chemical and genomic databases

Each skill includes the specific query formats, authentication methods, and parsing logic needed to interact with these databases effectively.

Technical Requirements

To use the Science Skills repository:

  • uv Python package manager is required (a fast Python package installer and resolver written in Rust)
  • Some skills require API keys for third-party services (AlphaGenome, OpenAlex, etc.)
  • The Apache 2.0 license permits commercial use, modification, and redistribution
  • The project is explicitly not an official Google product — it carries no support guarantees

Implications for Scientific Research

The release of DeepMind's Science Skills represents a step toward making AI agents practical tools for working scientists. Rather than requiring researchers to manually construct queries across dozens of different databases and services, a well-designed agent skill can orchestrate the entire workflow. As these skills mature and multiply, they could significantly accelerate the pace of scientific discovery by reducing the overhead of data access and analysis.

Key Takeaways

  • DeepMind released an open-source collection of modular AI agent skills for genomics, structural biology, cheminformatics, and literature search
  • Integrates with 30+ scientific databases including AlphaGenome, AlphaFold DB, UniProt, and OpenAlex
  • Requires the uv Python package manager; some skills need API keys
  • Licensed under Apache 2.0 — not an official Google product
  • Represents a step toward practical AI-assisted scientific research workflows