How to update the model

[hxpGit512]

Dear Celltypist Team,

If I have a new annotated dataset, how can I train and update it into the existing model ？

Looking forward to your reply！

[ChuanXu1]

@hxpGit512, you can check the Usage (Supplemental guidance -> generate a custom model) for details on how to train a model and save it into the existing model list. Also check out the celltypist.train function for the training parameters.

As an example, if you have an AnnData with highly variable genes, you can train the model using model = celltypist.train(adata[:, adata.var.highly_variable], 'cell_type_column', n_jobs = -1, max_iter = 500, check_expression = False) followed by model.write(f"{celltypist.models.models_path}/some_model_name.pkl")

[hxpGit512]

@ChuanXu1 , Multiple annotated datasets can be trained in batches and then incorporated into one model ？

[ChuanXu1]

@hxpGit512, do you mean multiple datasets with different annotations? If yes, you need to train and save each of them separately.