How to choose the best mode multi or full?

[vagm110901]

Hello,

I have tried to compare which, multi or full mode, is better to perform the deconvolution in my case, where I am working with Visium 10X Genomics Spatial Transcriptomics data of spinal cord with a single-nucleus reference. However, the mean result for the samples is quite disparate along the samples in some cell types.

Here you are an example of some spots. Why do neurons increase as much?

> listaObjAnnot[["reference"]]@assays[["full_mode"]]@data[,560]
       Astrcytes      Endothelial  Ependymal Cells      Lymphocytes         Meninges        Microglia          Neurons
    0.2264240292     0.1345648850     0.0002841117     0.1162233548     0.0002841117     0.0002841117     0.3234964723
             OPC Oligodendrocytes        Pericytes          Schwann
    0.0002841117     0.1286338291     0.0002841117     0.0692368713
> listaObjAnnot[["reference"]]@assays[["multi_mode"]]@data[,560]
       Astrcytes      Endothelial  Ependymal Cells      Lymphocytes         Meninges        Microglia          Neurons
       0.3895359        0.0000000        0.0000000        0.0000000        0.0000000        0.0000000        0.6104641
             OPC Oligodendrocytes        Pericytes          Schwann
       0.0000000        0.0000000        0.0000000        0.0000000

Or here. Why have I lost the Oligodendrocytes when they have a value similar to Astrocytes?

listaObjAnnot[["reference"]]@assays[["full_mode"]]@data[,16]
       Astrcytes      Endothelial  Ependymal Cells      Lymphocytes         Meninges        Microglia          Neurons
    0.1797289353     0.1294629853     0.1297238510     0.0002637652     0.0002637652     0.0002637652     0.3320902301
             OPC Oligodendrocytes        Pericytes          Schwann
    0.0002637652     0.1692216773     0.0002637652     0.0584534951
> listaObjAnnot[["reference"]]@assays[["multi_mode"]]@data[,16]
       Astrcytes      Endothelial  Ependymal Cells      Lymphocytes         Meninges        Microglia          Neurons
       0.4240144        0.0000000        0.0000000        0.0000000        0.0000000        0.0000000        0.5759856
             OPC Oligodendrocytes        Pericytes          Schwann
       0.0000000        0.0000000        0.0000000        0.0000000

I want to show also a comparison I have done for each cell type in both modes.

In the image I have represented the mean value of the weight for each cell type in all the spots from each slice. I have 4 slices from the same patient (reference & slice 2-4), 3 slices each one from a different sample (samples 2-4), and the mean for the seven slices.

What I am looking is that in general for the multi mode (blue) the weights are lower for most of the cell types. However I had understood that multi mode should be the four more abundant cell types per spot, what should mean a normalized value higher for all cell types and not only for the neurons in my case. I am not sure if maybe when the multi mode selects the four more abundant cell types, those weights from cell types which are not considered are added to the most abundant cell types increasing their weight and equaling the four of them.

I hope you could help me understanding and choosing the best option to my problem.
Thank you so much.

Victor Gaya