search

OmicsML / CellPLM

Official repo for CellPLM: Pre-training of Cell Language Model Beyond Single Cells.
BSD 2-Clause "Simplified" License
63 stars 6 forks source link

Is there a pre-trained cell type annotation model available with weights that can be directly used on the MS dataset? #16

Open noob000007 opened 1 month ago

noob000007 commented 1 month ago

Is there a pre-trained cell type annotation model available with weights that can be directly used on the MS dataset?

elif DATASET == 'MS': data_train = ad.read_h5ad(f'../data/c_data.h5ad') data_test = ad.read_h5ad(f'../data/filtered_ms_adata.h5ad') data_train.var = data_train.var.set_index('index_column') data_test.var = data_test.var.set_index('index_column') train_num = data_train.shape[0] data = ad.concat([data_train, data_test]) data.var_names_make_unique()

noob000007 commented 1 month ago

I'm comparing the performance of CellPLM and scGPT for cell type annotation on other datasets. scGPT provides a pre-trained model that can be directly used, but unfortunately, CellPLM does not offer a pre-trained model that can be used in the same way.

Therefore, I plan to use the cell_type_annotation.ipynb code to train on the DATASET == 'MS' and then compare the results. Is there a fairer method for comparison? I would appreciate it if you could provide a pre-trained model with weights for direct comparison as soon as possible. Thank you.