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

Interpretation of Prediction #47

Closed uddamvathanak closed 6 months ago

uddamvathanak commented 7 months ago

Hi author,

I have run the tutorial to test out on GEARS model and what I found is that the output of the prediction is not a gene expression in post perturbation and instead it is just a gene vector.

May I know how to obtained the post perturbation gene expression?

Thanks.

yhr91 commented 6 months ago

Thanks for your question. Can you share what you mean by a 'gene vector'? You could post the exact output in this thread.

GEARS outputs a vector of post-perturbation gene expression values

uddamvathanak commented 6 months ago

what I meant by gene vector is that the output is in the shape of (1, number of genes) instead of (number of cell, nunber of genes).

I would imagine that the output is for each cell instead like post perturbation gene expression in shape of (cell, gene)

Is my answer clear enough?

Thanks :D

yhr91 commented 6 months ago

Thanks for your question. The response should be in the shape of (number of perturbation categories, number of genes). This is because GEARS predicts the average perturbation effects for a specific perturbation category and does not predict perturbation effects at the level of individual cells.

uddamvathanak commented 6 months ago

Thanks it clear my doubt. Cheer :D