search

kaizhang / SnapATAC2

Single-cell epigenomics analysis tools
https://kzhang.org/SnapATAC2/
222 stars 24 forks source link

try different knn methods #264

Closed TingTingShao closed 6 months ago

TingTingShao commented 6 months ago

Hi,

I have already had a AnnDataSet:

AnnDataSet object with n_obs x n_vars = 5825 x 6062095 backed at 'microglia_7_500_50000.h5ads'
contains 3 AnnData objects with keys: '014h5ad_processed_features/D19-13184', '014h5ad_processed_features/D19-13179', '014h5ad_processed_features/D19-8607'
    obs: 'sample', 'leiden'
    var: 'count', 'selected'
    uns: 'AnnDataSet', 'macs3', 'reference_sequences', 'spectral_eigenvalue'
    obsm: 'X_spectral_harmony', 'X_umap', 'X_spectral_mnn', 'X_spectral'
    obsp: 'distances'

but I want to try different knn method, can I simply run

snap.pp.knn(datas, n_neighbors=50, use_rep="X_spectral_mnn", method="pynndescent") # to update the distance matrix based on batch corrected X_spectual_mnn matrix
snap.tl.leiden(datas) # recluster
snap.pl.umap(datas, color="leiden", interactive=False) # show cluster

Kind regards, tingting

kaizhang commented 6 months ago

Yes, you should see the differences between different runs.