search

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

Beginner question: GEARS on un-perturbed data #74

Closed bioinfonewguy closed 2 months ago

bioinfonewguy commented 2 months ago

Hello,

I am new to bioinformatics and these work flows in general, but I am attempting to learn by working through different tools.

I loved the GEARS paper, and I am trying to wrap my head around how it could be used.

Do I have to initially have a dataset which contains perturbations? Or could I pull a single celltype from a source like CellxGene, and perform in silico perturbations of interest? I know that's a beginner question, and apologize if it's too basic.

Thank you for your time!

yhr91 commented 2 months ago

Hi, thanks for your question. Unfortunately, the current version of GEARS has not been designed for training on unperturbed data. We recommend training on data from perturbed cells, where the cell type and experiment type of the training data match those that are desired at the time of inference.