Code repository for Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear baselines

This repository contains the code to reproduce our analysis from

Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear baselines. Constantin Ahlmann-Eltze, Wolfgang Huber, Simon Anders. Nature Methods 2025; doi: https://doi.org/10.1038/s41592-025-02772-6

A copy of the code is permanently archived at https://doi.org/10.5281/zenodo.14832393.

• The notebooks folder contains the R scripts used for the analysis and to make the figures
  • Double Perturbation Analysis
  • Single Perturbation Analysis
  • Dataset Overview
• The benchmark folder contains the scripts to reproduce the benchmark results
  • The benchmark/src contains individual scripts to run each method
  • The benchmark/conda_environments and benchmark/renv contain the details about the software versions
  • The benchmark/submission contains the script to launch the scripts using my custom workflow manager