# Robin: A multi-agent system for automating scientific discovery

**Authors:** Ali Essam Ghareeb, Benjamin Chang, Ludovico Mitchener, Angela Yiu, Caralyn J. Szostkiewicz, Jon M. Laurent, Muhammed T. Razzak, Andrew D. White, Michaela M. Hinks, Samuel G. Rodriques  
**arXiv:** [2505.13400](https://arxiv.org/abs/2505.13400)  
**Date:** 2025-05-19  
**Summary page:** https://www.summarizepaper.com/en/paper/2505.13400v1/

## Key points

- Scientific discovery involves background research, hypothesis generation, experimentation, and data analysis.
- Artificial intelligence has been increasingly used in scientific discovery but no system has fully automated all stages until now.
- Robin is the first multi-agent system capable of autonomously conducting key intellectual steps of the scientific process.
- Robin integrates literature search agents with data analysis agents to generate hypotheses, propose experiments, interpret results, and update hypotheses semi-autonomously.
- In a recent study focused on dAMD, Robin identified a novel treatment strategy by enhancing retinal pigment epithelium phagocytosis using ripasudil as a therapeutic candidate.
- Robin suggested and analyzed an RNA-seq experiment to understand how ripasudil upregulates phagocytosis by revealing the upregulation of ABCA1 as a potential new target.
- Robin can generate therapeutic candidates for experimental testing by querying Falcon based on scientific rationale and pharmacological profile.

## Summary

Scientific discovery is a complex process that involves background research, hypothesis generation, experimentation, and data analysis. While artificial intelligence has been increasingly used in scientific discovery, no system has fully automated all these stages in a single workflow until now. Introducing Robin, the first multi-agent system capable of autonomously conducting key intellectual steps of the scientific process. By integrating literature search agents with data analysis agents, Robin can generate hypotheses, propose experiments, interpret results, and update hypotheses in a semi-autonomous manner. In a recent study focused on , Robin identified a novel treatment strategy by enhancing retinal pigment epithelium phagocytosis. Through its autonomous processes, Robin proposed and validated ripasudil as a promising therapeutic candidate for dAMD. Ripasudil is a rho kinase (ROCK) inhibitor that had not previously been considered for treating dAMD. To further understand how ripasudil upregulates phagocytosis, Robin suggested and analyzed an RNA-seq experiment which revealed the upregulation of ABCA1, a critical lipid efflux pump and potential new target. Expanding on its capabilities, Robin can also generate therapeutic candidates for experimental testing by querying Falcon to evaluate each candidate based on scientific rationale and pharmacological profile. The ranked list of drug candidates can then be reviewed by human scientists for further testing in the lab. Overall, Robin represents a groundbreaking advancement in AI-driven scientific discovery by autonomously identifying and validating novel therapeutic candidates within an iterative lab-in-the-loop framework. Its ability to seamlessly integrate literature search with data analysis sets a new standard for AI-driven scientific research and opens up exciting possibilities for accelerating the pace of scientific discovery.