Robin: a multi-agent system for automating scientific discovery

A multi-agent system that integrates literature-search agents with data-analysis agents to automate hypothesis generation and data analysis for experimental biology. Applied to dry age-related macular degeneration, it proposed enhancing retinal pigment epithelium phagocytosis and identified and confirmed in vitro efficacy for two compounds, ripasudil and KL001.

L4Autonomous iterative research Source Peer-reviewed source Outcome Author-reported

At a glance

Type

Use case

Workflow level

L4 · Autonomous iterative research

Organisation

FutureHouse (with University of Oxford)

Year

2026

Disciplines

BiologyBiomedicine

Tasks

Data analysisDrug repurposingHypothesis generationTarget discovery

Entry updated

2026-10-15

Evidence & verification

Source level

Peer-reviewed source
Grounded in a peer-reviewed publication. This describes the source; it does not mean the reported results have been independently reproduced.

Outcome verification

Author-reported
Outcomes are reported by the team that built the system. The entry is source-checked against the team's own publications; not independently reproduced.

Sources checked

2026-10-15 (UTC)

Peer-reviewed in Nature (published online 2026-05-19, DOI 10.1038/s41586-026-10652-y, resolved 2026-10-15). In vitro validation was performed within the authors' own lab-in-the-loop workflow; no independent reproduction was identified at source-check time.

Source-checked does not mean independently reproduced. See the Method page for how these labels are assigned.

Sources (2)

Workflow & human oversight

Hypothesis generation and analysis iterate on experimental feedback. This is a bounded lab-in-the-loop workflow: humans execute wet-lab experiments, oversee safety and validate findings, not a fully autonomous laboratory.

Workflow level is not an evidence score; higher is not better evidence.

Problem

Discovery in biology proceeds through observation, hypothesis generation, experimentation and data analysis, and no system had yet automated all of these stages. The authors ask whether hypothesis generation and data analysis can be fully automated while experiments stay in a lab-in-the-loop.

Approach

Robin combines literature search agents with data analysis agents so it can generate hypotheses, propose experiments, interpret experimental results and generate updated hypotheses. When it proposed a mechanism follow-up, it also designed and analyzed the RNA-seq experiment that tested it.

Outcomes (as reported by the originating team)

The authors report that Robin identified and confirmed in vitro efficacy for ripasudil (a clinically used Rho kinase inhibitor never previously proposed for treating dAMD) and KL001, and that a follow-up RNA-seq analysis revealed upregulation of ABCA1, a lipid efflux pump and possible novel target. All hypotheses, experimental directions, data analyses and main-text figures were produced by Robin (Nature, 2026).

Limitations

Reported efficacy is in vitro only; no clinical trial or in vivo result is claimed. The workflow remains semi-autonomous: experiments are run by humans in the loop. All analyses and figures were produced by the system, so independent verification of the analysis pipeline is pending. Not independently reproduced.

Reproducibility

Code for Robin is published at the linked repository (link verified live 2026-10-15). Reproducing the discovery claim itself would require the wet-lab validation pipeline.

Adoption requirements

A working molecular-biology wet lab for validation; access to the system's code and foundation models; researchers willing to act as the experimental loop and validate system-proposed experiments.

Comments

No comments yet.

Add a comment

Comments are held for moderation before appearing. Critical or corrective comments are welcome — especially evidence that an entry overstates what has been verified.