Testing a simple linear baseline on the virtual cell challenge.
The baseline comes from this publication claiming simple linear baselines still outperform deep-learning models. This baseline ranked 1,022 in VCC 2025.
Ahlmann-Eltze, Constantin, Wolfgang Huber, and Simon Anders. "Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines." Nature Methods (2025): 1-5.