Inquiry about location of ground truth datasets

[SGNYYYY]

Hello! Thank you for sharing this valuable project.

While trying to run the code, I noticed in the run_multi_modality() function there's a specified path for the ground truth data:

adata_source_dir = "/mnt/volume3/trn/10_spatialMETA_revision/06_spatialmeta_groundtruth"

I've checked the project documentation and the dataset links you provided, but I couldn't find these ground truth datasets. Could you please clarify:

1. Will these ground truth datasets be made available in a future update?
2. Is there a specific way to access them or permissions that need to be requested?
3. If there are alternative datasets we could use for testing, could you provide guidance on that?

Thank you very much for your assistance! I look forward to your response.

[CHENyiru3]

Hi, I think this path is not correctly removed from code, so this doesn't mean that there should be a dataset or a actual folder for this?

You can try to find out datasets in article. They are all public datasets I believe, and it works well for me: Data availability
The combined ST and SM data, along with the scRNA-seq data for ccRCC processed in this study, have been deposited in the Zenodo database under accession code (https://zenodo.org/records/14986870). The combined ST and SM data for GBM are available in the Dryad database under accession code (https://doi.org/10.5061/dryad.h70rxwdmj), and the combined ST and SM data for the mouse brain are available at Mendeley Data under accession code (https://doi.org/10.17632/w7nw4km7xd.1). The processed datasets generated by SpatialMETA are available at Zenodo under accession code (https://zenodo.org/records/12528191). Source data for all Figures and Supplementary Figures are provided with this paper. Source data are provided with this paper.

Hope this can help you!

[RuonanTian]

I have upload the ground truth adata, https://zenodo.org/records/18228912. The adata.obs['pathological_annotation'] is the ground truth used in this study

[SGNYYYY]

I have upload the ground truth adata, https://zenodo.org/records/18228912. The adata.obs['pathological_annotation'] is the ground truth used in this study

Thank you so much for sharing the ground truth adata file and the Zenodo link! This is exactly what I needed.

[deadblackhole]

Hi, I think this path is not correctly removed from code, so this doesn't mean that there should be a dataset or a actual folder for this?

You can try to find out datasets in article. They are all public datasets I believe, and it works well for me: Data availability The combined ST and SM data, along with the scRNA-seq data for ccRCC processed in this study, have been deposited in the Zenodo database under accession code (https://zenodo.org/records/14986870). The combined ST and SM data for GBM are available in the Dryad database under accession code (https://doi.org/10.5061/dryad.h70rxwdmj), and the combined ST and SM data for the mouse brain are available at Mendeley Data under accession code (https://doi.org/10.17632/w7nw4km7xd.1). The processed datasets generated by SpatialMETA are available at Zenodo under accession code (https://zenodo.org/records/12528191). Source data for all Figures and Supplementary Figures are provided with this paper. Source data are provided with this paper.

Hope this can help you!

Thank you for your detailed reply and the shared dataset links – they have been incredibly helpful !

However, I’ve encountered a naming inconsistency between the sample identifiers in the paper and the raw filenames in the downloaded datasets, and I wonder if you could help clarify the mapping:

1. Mouse brain dataset

In Supplementary Data 2 (Excel file) of the paper, the mouse brain sample names are listed as m1_FMP, m3_FMP, m4_FMP.
But in the datasets downloaded from Mendeley Data (https://doi.org/10.17632/w7nw4km7xd.1):

• The filenames in the msi folder are ITO1, ITO2, ITO3
• The filenames in the visium folder are V11T17-101_A1, V11T17-101_B1, V11T17-101_C1, V11T17-101_D1

2. GBM dataset

In the paper, the sample is referenced as 248_T.
In the Dryad dataset (https://doi.org/10.5061/dryad.h70rxwdmj):

• Only 248_T_ST is accurately labeled for the ST data

Could you please inform me of two key points to facilitate my analysis:
The exact mapping between the mouse brain sample names in the paper (i.e., m1_FMP, m3_FMP, m4_FMP) and the raw filenames in the Mendeley Data dataset (e.g., ITO1/ITO2/ITO3 in the msi folder, V11T17-101_A1/B1/C1/D1 in the visium folder)?
How to correctly extract or identify the SM data corresponding to the 248_T sample from the GBM dataset on Dryad? The SM section shows that multiple samples (e.g., #UKF248_T, #UKF259_T) are stored together without clear labels indicating which part corresponds to 248_T.
Your clarification on these two points will be crucial for me. Thank you for your time and help!

[CHENyiru3]

Oh I am sorry, I don't know the precise steps in preprocessing. So I can't tell how to map the names in article to the original data... But here I can provide a trick, hope this may solve this issue: you can check the shape of your interested slice in SpatialMETA article, and then go to the potential original article to check their images, and you can know which one is the related raw data.

Maybe this is not the direct answer, but hope this can help you.

[deadblackhole]

Thanks for your reply; I will try this method.