Extracting Normalized Expression from Trained Model on Variable Subset

Help scvi-tools
1

Hello, apologies for a somewhat trivial question.

I am trying to extract the normalized expression matrix from my trained model, which was done so on a subset of variable genes. I want to plot a heatmap of the expression but not all of my curated markers appear in the variable genes list used to train the model. I was under the impression I could still extract normalized expression for any gene regardless of whether it was used to train the model. Is this correct or do I need to retrain the model including genes of interest to extract their denoised-normalized expression for visualization?

Thanks in advance!

2

Hi, scVI models can only generate gene expression for genes used during training, so yes you would have to retrain your model.

3

Okay yes I thought so, just wanted to confirm. Thanks for that!