Pretrained models on Tabula sapiens

Hello scvi-tools team,

I am wondering if there are any pretrained models(using let’s say SCVI) trained on taking entire Tabula sapiens as the reference dataset. It will be great if anyone can provide me with it.


We don’t have plans to train a single model on the entire TS dataset. We strive instead to train models on separate tissues (knowledge is transferrable to query datasets using tools such as scArches). We currently have pre-trained scvi models on a few of the TS datasets. If you’re interested, I can point you to those.

Yeah sure. It would be a great help. Also, is there any particular reason for not training on the entire TS dataset. I know the dataset is huge and may incur computation cost but is there any other reason apart from that.

Thank you.

It is still work on progress. But you can find what we have so far here on Zenodo (for the models, and the datasets are here).
Re training on the entire dataset, compute cost is the main reason I can think of.

Thanks a lot for that.