Supporting the analysis of multi-sample SRT data

[Antidington]

Dear author,

Thank you for developing these excellent tools.

I currently have dozens of SRT samples that are unstructured and highly heterogeneous. I would like to know if Hotspot is capable of identifying cross-sample gene expression patterns?

Looking forward to your reply!

Best regards,
Jichen

[Li-ZhiD]

Same problem! thanks!

[yuanjianwen-gif]

Any solutions? Thanks！

[Antidington]

Hi @Li-ZhiD @yuanjianwen-gif,

Here's my current approach for multi-sample SRT data:

I perform SCTransform on each sample individually to remove batch effects (depth and MT%), then use the SCT matrices as input for Hotspot. Since Hotspot doesn't explicitly use spatial coordinates, I treat this similar to a standard single-cell workflow.

However, I'm not entirely confident this is the optimal approach. Would you have any suggestions or alternative strategies for handling multi-sample SRT data with Hotspot?

Thanks!

[yuanjianwen-gif]

Hi!
Glad to hear you.
1. Using SCT could be a good idea, but shell we merge all samples then run SCT if we want to remove batch effects?
2. https://www.biorxiv.org/content/10.1101/2024.10.21.619529v1 this biorxiv paper used Jaccard index to select and merge(?) modules with several slides. I think there could be one way but I have not try, better to get more cases as referrence.

Hi @Li-ZhiD @yuanjianwen-gif,

Here's my current approach for multi-sample SRT data:

I perform SCTransform on each sample individually to remove batch effects (depth and MT%), then use the SCT matrices as input for Hotspot. Since Hotspot doesn't explicitly use spatial coordinates, I treat this similar to a standard single-cell workflow.

However, I'm not entirely confident this is the optimal approach. Would you have any suggestions or alternative strategies for handling multi-sample SRT data with Hotspot?

Thanks!

[Antidington]

Hi! Glad to hear you. 1. Using SCT could be a good idea, but shell we merge all samples then run SCT if we want to remove batch effects? 2. [https://www.biorxiv.org/content/10.1101/2024.10.21.619529v1 ](https://www.biorxiv.org/content/10.1101/2024.10.21.619529v1) this biorxiv paper used Jaccard index to select and merge(?) modules with several slides. I think there could be one way but I have not try, better to get more cases as referrence.

Hi @Li-ZhiD @yuanjianwen-gif,
Here's my current approach for multi-sample SRT data:
I perform SCTransform on each sample individually to remove batch effects (depth and MT%), then use the SCT matrices as input for Hotspot. Since Hotspot doesn't explicitly use spatial coordinates, I treat this similar to a standard single-cell workflow.
However, I'm not entirely confident this is the optimal approach. Would you have any suggestions or alternative strategies for handling multi-sample SRT data with Hotspot?
Thanks!

Hi!

Thank you for sharing these valuable insights!

1. Based on the discussion, I believe performing SCTransform on each sample individually is the preferred approach, even when aiming to remove batch effects.
2. Running Hotspot individually and calculating Jaccard index for module selection across multiple slides is an excellent point, but it seems more suitable for same disease state datasets (e.g., all tumor samples). In my case, I want to explore the differences between case and control which are highly heterogeneous. I think the way to calculate Jaccard index may underestimate the real situation.
3. In the end, I also want to pin an article. This article performed Hotspot on the whole dataset and identified robust modules through quantitative metrics combined with manual curation. I think this approach is more suitable, but they didn't test it on spatial data, only on single-cell data.

[yuanjianwen-gif]

Great! "explore the differences between case and control which are highly heterogeneous" is very common. I would like to give a try!