Reproduce “Towards Universal Gene Regulatory Network Inference: Unlocking Generalizable Regulatory Knowledge in Single-cell Foundation Models” 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

Jiaxin Qi, Hang Li, Yan Cui, Yuhua Zheng, Jianqiang Huang

This research introduces a framework and advanced methods to extract generalizable gene regulatory network (GRN) knowledge from pre-trained single-cell Foundation Models (scFMs). It achieves an average AUPRC of 0.768 for universal GRN inference on scBenchmark, establishing a new state-of-the-art for predicting regulatory links across unseen genes and diverse datasets.

Key insights

  • Single-cell Foundation Models (scFMs) contain rich, generalizable biological priors for gene regulation, but their potential is unlocked only through advanced querying mechanisms, not simplistic perturbation or attention interpretation.
  • Decoupling feature extraction from observed dataset-specific statistics via virtual queries (VVP and GDT) allows for the discovery of context-independent and universally applicable gene regulatory knowledge.
  • Regulatory relationships can be effectively characterized by analyzing both value interval responses (VVP) and instantaneous local sensitivities (GDT) within the latent space of frozen scFMs.

Summary by alphaXiv.

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