Dear Author,
Your work is excellent. I am writing to ask for some guidance on reproducing the gene expression prediction section of your paper.
I have been attempting to replicate your results using the heart data from [Heart Cell Atlas] (www.heartcellatlas.org), which is consistent with the example data provided in your repository.
I noticed that in your validation set and provided examples, you utilized HCAHeartST11702009.h5ad. Initially, I attempted to apply the method to a different dataset, HCAHeartST11702010.h5ad, but I was unable to achieve satisfactory results.
Upon further investigation, I believe the issue stems from the calculation of the similarity matrix.
When I tried to reproduce the results using the original dataset (HCAHeartST11702009.h5ad) but calculated the similarity matrix from scratch (following the "basic usage" preprocessing steps instead of using your pre-computed matrix), I still could not reproduce the high performance demonstrated in your work.
It seems that the similarity matrix I generated during preprocessing is not as effective as the one provided in your example.
Could you please offer some guidance on the specific details of calculating the similarity matrix? I would greatly appreciate it if you could clarify:
Preprocessing Parameters: What specific parameters were used for segmentation and encoding?
Matching/Alignment: How did you ensure correct matching during this process?
Are there any additional steps in the pipeline that differ from the standard basic usage tutorial?
Thank you very much for your time and help!
Best regards,
Yuanchao Liu
Dear Author,
Your work is excellent. I am writing to ask for some guidance on reproducing the gene expression prediction section of your paper.
I have been attempting to replicate your results using the heart data from [Heart Cell Atlas] (www.heartcellatlas.org), which is consistent with the example data provided in your repository.
I noticed that in your validation set and provided examples, you utilized HCAHeartST11702009.h5ad. Initially, I attempted to apply the method to a different dataset, HCAHeartST11702010.h5ad, but I was unable to achieve satisfactory results.
Upon further investigation, I believe the issue stems from the calculation of the similarity matrix.
When I tried to reproduce the results using the original dataset (HCAHeartST11702009.h5ad) but calculated the similarity matrix from scratch (following the "basic usage" preprocessing steps instead of using your pre-computed matrix), I still could not reproduce the high performance demonstrated in your work.
It seems that the similarity matrix I generated during preprocessing is not as effective as the one provided in your example.
Could you please offer some guidance on the specific details of calculating the similarity matrix? I would greatly appreciate it if you could clarify:
Preprocessing Parameters: What specific parameters were used for segmentation and encoding?
Matching/Alignment: How did you ensure correct matching during this process?
Are there any additional steps in the pipeline that differ from the standard basic usage tutorial?
Thank you very much for your time and help!
Best regards,
Yuanchao Liu