Reproduce “CellCanon: An Agentic Cellular World Model for Prediction, Explanation, and Observation-Driven Updating” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zhaokang Liang, Shuyang Zhuang, Xiaoran Jiao, Hao Chen

CellCanon predicts missing drug–cell gene responses by compiling source-linked evidence once per condition and sharing it across gene-specific numerical readouts. It links predictions to claims. Across 339 held-out SciPlex3 conditions in three cell lines, its expression-derived DE AUPRC is 0.8067, versus 0.4938 for VCWorld. The test covers 24-hour pooled-dose responses; updates reveal existing measurements in simulation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.