Perturbation modeling aims to find a model that can accurately predict the outcome of a perturbation on a given cell, such as a drug or CRISPRi intervention. Creating a massive amount of perturbation data is a challenging endeavor due to technical and financial constraints, which inhibit the progress of creating a biologically accurate virtual cell. Cell foundation models

promise a remedy to this issue: by pretraining a model with abundant unperturbed single-cell transcriptomic data one can, in theory, increase the performance of perturbation models trained on limited data.