Visium deconvolution reference datasets

Question

[Claire Watson]

I am investigating some methods for Visium deconvolution. Looked at some benchmark papers and most of the top accurate algorithms relied heavily on reference datasets.

So can anyone suggest me source to find trustable - datasets? I tried to explore cellxgene but seem like their metadata is messy and not consistent.

Answer 1

[GenoMax]

Looks like 10x provides example datasets (you will need to register to access) : https://www.10xgenomics.com/resources/datasets?query=&page=1&configure%5BhitsPerPage%5D=50&configure%5BmaxValuesPerFacet%5D=1000&refinementList%5Bproduct.name%5D%5B0%5D=Spatial%20Gene%20Expression

Visium discovery hub: https://www.10xgenomics.com/research-areas/cancer/visium-discovery-hub

[Claire Watson]

Oh Great! Thank you so much. I'll definitely check it out!

Answer 2

[jackson9]

True, it's the reference dataset that determine result of deconvolution, not the algorithm.

A good reference datasets depend on lots of factors such as tissue, condition, cell types, unit... related to your visium dataset. So you have to filter out yourself, unfortunately.

[jackson9]

It's worth to take a look at BioTuring. This company is well-known with its standardized metadata, so I think they have lots of good reference scRNA datasets there.

They also implemented cell2location in their app. You can book a demo to see how it works