Possible to use multiple single cell datasets in Tangram? about tangram HOT 3 OPEN

Hi @boyangzhang1993,

Thank you for using Tangram!

The recommendation for the Tangram is to use single cell data and spatial data coming from the same tissue for the most meaningful results. However, we have used Tangram in situations where single cell and spatial data come from different tissue sections but ideally from the same species and tissue type.

While the mapping result will vary if you're using a combined dataset, I think using a combined single cell dataset will help in predicting gene expressions (incase of gene imputation). So depending on your use case:

1. If you're interested in mapping single cell data onto space, I would try mapping the single cell data individually to the spatial data and compare the results.
2. However, if you're interested in predicting/correcting the expression of genes in spatial data, I would recommend using the combined dataset.

from tangram.

Thank you for your detailed response regarding the use of Tangram with single-cell datasets.

In my specific case, I have multiple single-cell datasets representing liver tissue with cancer and liver tissue without cancer, and spatial transcriptomic data. My primary goal is to use Tangram to predict whether a specific voxel in the spatial data represents cancerous or non-cancerous liver cells.

Given this context:

Would it be more effective to combine the datasets (both cancerous and non-cancerous) to create a comprehensive representation of the liver tissue and then proceed with the prediction? Or,

Would it be advisable to process the cancerous and non-cancerous datasets separately, predicting for each, and then comparing the outcomes? It seems that the results from each round of Tangram are not comparable. Is that right?

from tangram.

Hi @boyangzhang1993,

In your case, since you want to be able to analyze if a voxel has cancerous/ non-cancerous cells, combining single cell data with cancerous and non-cancerous cells would help!

from tangram.