SPHINX: An Agentic AI System for Automated Scientific Discovery in Computational Biology

The rapid growth of biomedical data and analysis techniques has created a critical research bottleneck, leading to inefficient discovery workflows. We introduce SPHINX (Scientific Progressive Hypothesis Iteration and Organized Novel In-silico eXperiments), a multi-agent AI system designed to accelerate discovery by automating computational biology. SPHINX integrates specialized agents for literature synthesis, hypothesis generation, data analysis and visualization, and in-silico experimentation within a “Lab-in-the-Loop” framework. Through rigorous case studies in Alzheimer's Disease and immunology, SPHINX replicated findings and performed complex analyses by autonomously reasoning, writing code, and self-correcting errors, demonstrating its end-to-end capability to accelerate biomedical research.