How to create Giotto object with precomputed results?

1. The Giotto class

A Giotto object or gobject can be created with createGiottoObject or createGiottoVisiumObject for Visium datasets.

gobject = createGiottoObject()

gvisiumobject = createGiottoVisiumObject()

createGiottoVisiumObject expects a typical Visium 10X data folder structure or a path to the h5 file, but createGiottoObject is more flexible and can also accept precomputed results (e.g. PCA, UMAP, network, …)

2. Giotto slots

2.1 Required slots

The minimum requirements to make a Giotto class object are:
- raw_exprs: matrix with raw expression values
- spatial_locs: spatial coordinates for the cell (or spot) centroids. If this is not provided it will actually create dummy coordinates.

2.2 Additional matrix slots

• norm_expr: slot to keep matrix with normalized expression values
  • use expression_values = ‘normalized’ to access this slot in downstream functions
    
• norm_scaled_expr: slot to keep matrix with normalized and scaled expression values
  • use expression_values = ‘scaled’ to access this slot in downstream functions
    
• custom_expr: slot to keep matrix with custom values
  • use expression_values = ‘custom’ to access this slot in downstream functions

2.3 Metadata slots

• cell_metadata:
  
• gene_metadata:

2.4 spatial information slots

• spatial_network:
  
• spatial_network_name:
  
• spatial_grid:
  
• spatial_grid_name:

2.5 spatial enrichment slots

• spatial_enrichment:
  
• spatial_enrichment_name:

2.6 dimension reduction slots

• dimension_reduction: accepts list of dimension reduction objects

Example to add your own pre-computed PCA results

# step1: create matrix from PCA coordinates and name the columns (PCs): Dim.1, Dim.2, Dim.3, ...
# e.g. PCA_matrix


# step2: create a pca dimension reduction object
PCA_list <- list(
     'type' = "cells", # only PCA on cells is currently supported
     'name' = "pca",   # name of your pca, 'pca' is the default name
     'reduction_method' = "pca", # needs to be pca
     'coordinates' = as.matrix(PCA_matrix), # make sure to provide a matrix
     'misc' = NULL) # slot for miscellaneous data, e.g. list(eigenvalues = ..., loadings = ...)


# step3a: add PCA_list to createGiottoObject
gobj <- createGiottoObject(...,
                           dimension_reduction=list(PCA_list), ...)

# step3b: add multiple PCA_lists to createGiottoObject
# the name parameter should be different for each PCA list
gobj <- createGiottoObject(...,
                           dimension_reduction=list(PCA_list1, PCA_list2, ...))

# step4: for UMAP or tSNE:
# same as for PCA, except for UMAP: reduction_method = 'umap' and default name = 'umap'}
# same as for PCA, except for UMAP: reduction_method = 'umap' and default name = 'umap'}

2.7 nearest neighbor network slots

• nn_network:

2.8 slot for images

• images:

2.9 slot for offset file

• offset_file:

2.10 slot for instructions

• instructions: