jianhuupenn / SpaGCN

SpaGCN: Integrating gene expression, spatial location and histology to identify spatial domains and spatially variable genes by graph convolutional network
MIT License
199 stars 60 forks source link

Inconsistent results in SVGs identification compared to tutorial.ipynb #69

Open aleksa1902 opened 1 year ago

aleksa1902 commented 1 year ago

Dear Dr. Hu,

I followed the tutorial.ipynb to identify SVGs in my dataset. However, I noticed that I am getting only 2 SVG solutions instead of the expected 5 as mentioned in the tutorial. A similar issue occurs when identifying meta genes.

Here are the steps: Preprocessed the data according to the tutorial instructions. Ran the SVG identification code from tutorial.ipynb. Expected to obtain 5 SVG solutions, but only received 2. I also noticed that when I run the tutorial in ez mode, the results are consistent with the expected solutions. It seems that there might be a discrepancy in the result.h5ad file used in tutorial.ipynb compared to the one provided for download.

I have attached a screenshot of the code and the obtained results for reference. Any insights into why I am experiencing these discrepancies and suggestions for resolving the issue would be greatly appreciated.

Screenshot: image

Thank you for your help!