search

jianhuupenn / SpaGCN

SpaGCN: Integrating gene expression, spatial location and histology to identify spatial domains and spatially variable genes by graph convolutional network
MIT License
200 stars 59 forks source link

problem reproducing turtorial #65

Open watergua opened 1 year ago

watergua commented 1 year ago

Hi, I was running the turtorial notebooks on my computer. Everthing seemed alright, the calculation of l and res returned same value as the turtoral. But there was a warning when identifying SVGs: WARNING: It seems you use rank_genes_groups on the raw count data. Please logarithmize your data before calling rank_genes_groups. The resrult image of spatial domains looked very different, the clusters are different and the spatial organization is bad. SVGs of cluster 0 can not be captured, and for my clustser 1 is also different SVGs for domain 1 : ['MT-ND2', 'CCK', 'MT-ND3', 'CAMK2N1', 'ENC1', 'MT-ND5', 'ARPP19', 'HOPX', 'HPCAL1']. This happened for both notebooks tutorial.ipynb and tutorial_ez_mode.ipynb. image Anyone has an idea on what is wrong and how to debug?

KalinNonchev commented 1 year ago

I also have troubles reproducing the exact picture. In the tutorial they use spagcn 1.2.0 and it is not available anymore. I use 1.2.5 and this is the result I obtain:

Screenshot 2023-03-27 at 13 14 34
swainasish commented 1 year ago

I also followed the steps mentioned in the tutorial script but got the result differ from what mentioned in the paper for libd 151673. image