Reproduce “BEND 2.0: Benchmarking the Capabilities Needed to Design DNA” 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

Andreas Bjerregaard, Rachael DeVries, Dionysia Danai Brilli, Panagiotis Antoniadis, Yan Li, Anders Krogh, Wouter Boomsma, Ole Winther and 1 more

BEND 2.0 evaluates DNA models on predicting regulatory effects and ranking edits for responses. It tests function, cellular context, mutation interactions, and design using eight tasks across five experimental datasets. On a target-matching task, AlphaGenome’s top ten included a matching response for 62% of 331 targets, versus 38% with uniform selection; this concerns assayed libraries, not arbitrary designs.

Summary by alphaXiv.

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