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

KeyError: 'val' #71

Closed xiaomaolxc closed 1 month ago

xiaomaolxc commented 3 months ago

I am interested in simulating perturbations for a specific gene, as mentioned in your article, but I have encountered some issues as shown in this link: https://github.com/snap-stanford/GEARS/issues/56. As described in your article about simulating single-gene perturbations, I would like to know how you handled individual perturbation categories. How did you manage single perturbations in your study?

yhr91 commented 3 months ago

Sorry, I don't follow your question - can you please be more specific on what errors you are facing. What do you mean by 'manage' single perturbations?

xiaomaolxc commented 3 months ago

I'm glad for your help. My idea is to use this tool to simulate perturbations of a specific gene. As shown below: The condition column in my ada only contains two situations: AAACCCAAGATGACCG_1-0 ctrl AAACCCAGTCAAATCC_1-0 ctrl Name: condition, dtype: category Categories (2, object): ['CHRNB1+ctrl ',' ctrl '] 微信图片_20240604102855

xiaomaolxc commented 3 months ago

I'm glad for your help. My idea is to use this tool to simulate perturbations of a specific gene. As shown below: The condition column in my adata only contains two situations: AAACCCAAGATGACCG_1-0 ctrl AAACCCAGTCAAATCC_1-0 ctrl Name: condition, dtype: category Categories (2, object): ['CHRNB1+ctrl ',' ctrl '] [cid:f1d8e530-bf45-41b0-bcdf-a401e70d6f8b]

From: Yusuf Roohani @.> Sent: Sunday, June 2, 2024 08:10 To: snap-stanford/GEARS @.> Cc: xiaomao Liu @.>; Author @.> Subject: Re: [snap-stanford/GEARS] KeyError: 'val' (Issue #71)

Sorry, I don't follow your question - can you please be more specific on what errors you are facing. What do you mean by 'manage' single perturbations?

— Reply to this email directly, view it on GitHubhttps://github.com/snap-stanford/GEARS/issues/71#issuecomment-2143744107, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BBABJ5LIZSBA6ERK5DSD3JTZFLHQ5AVCNFSM6AAAAABISYQSWWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNBTG42DIMJQG4. You are receiving this because you authored the thread.Message ID: @.***>

yhr91 commented 1 month ago

Sorry for the late reply, I was busy with my PhD defense.

As described in https://github.com/snap-stanford/GEARS/issues/56, those are too few perturbation categories for the model to perform the default 'simulation' split. I also wouldn't recommend training GEARS on only a single perturbation under any split.

Please feel free to re-open this issue in case you still have trouble, and please add more context on the exact error you are facing.