snap-stanford / GEARS

GEARS is a geometric deep learning model that predicts outcomes of novel multi-gene perturbations
MIT License
189 stars 38 forks source link

prediction interpretation #46

Closed Gianl-msi closed 6 months ago

Gianl-msi commented 7 months ago

The prediction tutorial notebook generates values between 0 and 5. Are those absolute expression changes? Or expression values post perturbation? If those are the final expression value, what's the fastest way to retrieve the unperturbed values?

yhr91 commented 7 months ago

Thanks for your question. The output of the predict function is expression values post perturbation.

To retrieve unperturbed values you can simply call gears_model.ctrl_expression for average unperturbed state or gears_model.ctrl_adata for gene expression values for all control cells from the original dataset.