maxpmx / histogene Goto Github PK

Hi there,

I notice that skin_hvg_cut_1000.npy is missing when I try to run the ViT_SKIN dataset.

The cscc dataset files have been downloaded correctly. I have tried to use the preprocess functions in utils.py, and scanpy package functions like highly_variable_genes, filter_genes according to the description of preprocess in the paper:

We then select the top 1,000 highly variable genes in each tissue section and eliminate genes that are expressed in less than 1,000 spots across all tissue sections.

like this:

adata = ad.concat(adata_list, merge='only')
adata.obs_names_make_unique()
adata.var_names_make_unique()
sc.pp.normalize_total(adata)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, n_top_genes=1000, subset=True)
sc.pp.filter_genes(adata, min_counts=1000)

but the result is different from 134 genes remained according to Supplementary Note 1.

So can we have the real preprocess function that can generate skin_hvg_cut_1000.npy?

Thank you!

Thank you for providing such excellent work. We would like to use the model for some analytical tasks. Could you please provide parameters for other sliced models? We appreciate your help very much.

Hi there! Thanks a lot for the great work!
I am currently applying your method on my dataset. May I know how large the training and validation loss were during your experiments? Thanks!

Hello!

I find that When I try to realize ST-Net and calculate its ARI on this dataset, the result is different from what the paper describes, though I have applied the default hyperparameters mentioned in the ST-Net paper.

So could you offer your trained ST-Net, or detailed hyperparameters you used to train (such as backbone(DenseNet or ResNet), epochs, learning rate, optimizer and so on)?

Thank you!