OpenAI launches Rosalind Workbench, a workspace for biological research in ChatGPT

OpenAI launches Rosalind Workbench, a workspace for biological research in ChatGPT

On 28 August, OpenAI released a preview of Rosalind Workbench, a workspace that supports biological research from data and specialized tools through analysis planning and verifiable results.

Biological data, analytical work, and records of how results were obtained often reside in separate services. In April, OpenAI introduced GPT-Rosalind, a specialized model for biological research and drug development. Rosalind Workbench brings the research question, data, tool actions, experimental plan, and intermediate results together in one place, allowing researchers to see how each step led to the next.

OpenAI illustrates this process with a question about disease mechanism. A signal from genetic data can be examined alongside a molecular structure, published research, and an experimental design. In a sequencing workflow, which involves reading DNA or RNA sequences, Workbench links data files to sample information, checks data quality, helps select independent samples for comparison, and defines the statistical analysis plan.

Workbench has two modes. Explore mode uses the ChatGPT models available to the user to answer general scientific questions. Research mode is designed for complex biological questions and in-depth analysis. Verified members of organizations can request access on behalf of their organization, while OpenAI plans to provide individual access later.

In the workflow shown by OpenAI, Rosalind prepares a plan, the researcher approves it, and Rosalind then coordinates the selected tools. The results and the supporting data remain linked to the plan, allowing the researcher to trace and verify the workflow before making the next decision.