Robin: The AI System Automating Scientific Discovery, Unveils New Blindness Treatments

Scientific discovery, traditionally a meticulous and iterative process involving observation, hypothesis generation, experimentation, and data analysis, has long been a domain reliant on human intellect and intuition. While artificial intelligence has made significant strides in various biological applications, no single system has previously managed to automate all these critical stages comprehensively. This landscape has now been transformed with the introduction of Robin, a groundbreaking multi-agent AI system that marks a pivotal moment in the automation of scientific research. Developed by a team including Ali Essam Ghareeb and Benjamin Chang, Robin stands as the first system capable of fully automating both hypothesis generation and data analysis specifically for experimental biology. By seamlessly integrating specialized literature search agents with sophisticated data analysis agents, Robin can autonomously generate novel hypotheses, propose detailed experimental designs, interpret complex experimental results, and even formulate updated hypotheses based on new findings. This semi-autonomous approach significantly accelerates the pace of discovery, moving beyond the fragmented AI applications seen before. One of Robin's most compelling applications to date has been in the realm of ophthalmology, specifically targeting dry age-related macular degeneration (dAMD). This condition is recognized as the leading cause of blindness in developed nations, posing a significant global health challenge. Through its iterative "lab-in-the-loop" framework, Robin successfully identified promising therapeutic candidates for dAMD, demonstrating its practical efficacy in addressing complex medical problems. The system's discoveries for dAMD are particularly noteworthy. Robin proposed enhancing retinal pigment epithelium phagocytosis as a novel therapeutic strategy. Following this, it identified and confirmed the in vitro efficacy of two compounds: ripasudil and KL001. Remarkably, ripasudil, a clinically used Rho kinase (ROCK) inhibitor, had never before been considered or proposed for the treatment of dAMD. To further unravel the mechanisms behind ripasudil's action, Robin independently proposed and analyzed a follow-up RNA-seq experiment, which subsequently revealed the upregulation of ABCA1, a lipid efflux pump, suggesting a possible novel therapeutic target. What truly underscores Robin's transformative potential is that every hypothesis, experimental direction, data analysis, and even the data figures presented in the main text of the research report were entirely generated by the AI system itself. This unprecedented level of autonomy establishes Robin as the first AI system to autonomously discover and validate novel therapeutic candidates within an iterative framework. Its introduction not only streamlines the scientific process but also ushers in a new paradigm for AI-driven scientific discovery, promising to revolutionize how we approach complex research challenges across various disciplines.