Claude for Life Sciences Research

Anthropic aims to help researchers make new discoveries, and further, to enable AI models to autonomously carry out discoveries. Previously used for individual tasks such as writing statistical analysis code or summarizing papers, it now aims to support the entire process from initial discovery to commercialization.

Claude Sonnet 4.5 shows superior performance on life sciences-related tasks compared to previous models. On the Protocol QA benchmark, which measures understanding of experimental protocols, it scored 0.83, surpassing the human baseline (0.79), and it also demonstrated improved performance on BixBench, an evaluation of bioinformatics tasks.

To increase connectivity with scientific tools, the following new Connectors have been added:

• Benchling: Provides links to experiment and record data
• BioRender: Connects scientific diagrams and template libraries
• PubMed: Access to millions of biomedical research papers
• Wiley Scholar Gateway: Provides verified academic content
• Synapse.org: Supports data sharing and collaborative analysis
• 10x Genomics: Performs single-cell and spatial analysis via natural language

In addition, Agent Skills have been introduced to consistently perform specific protocols. A representative example is the single-cell-rna-qc skill, which follows scverse best practices to perform quality control and filtering of single-cell RNA sequencing data.