Build a K-Nearest Neighbors (KNN) Graph and Store in Seurat Object

This function constructs a K-Nearest Neighbors (KNN) graph based on selected principal components (PCs) from a Seurat object. The resulting graph and associated settings are stored in the misc slot of the Seurat object under NNet.setting.

Usage

prepare.graph(seurat.obj, knn = 30)

Arguments

seurat.obj: A Seurat object prepared using prepare.seurat, which includes PCA results in its NNet.setting.
knn: An integer specifying the number of nearest neighbors to use for building the KNN graph. Default is 30.

Value

A Seurat object with its NNet.setting updated to include:

p: A sparse cell-by-cell affinity matrix.
nn.idx: A cell-by-neighbour matrix where each row contains the indices of the nearest neighbors for the corresponding cell.
nn.w: A matrix describing weights of connections between neighbouring cells described by nn.idx

Details

The function constructs a KNN graph using the most informative PCs learned during the execution of prepare.seurat. The graph is built with the build.graph function, which computes a sparse affinity matrix, nearest neighbor indices, and weights for each cell. These results are stored in the Seurat object's NNet.setting, which is used for downstream co-expression analysis, including PC regression and network inference.

The KNN graph is a critical structure that captures relationships between cells in the reduced PC space, enabling scalable and biologically meaningful analyses.

Examples

# Select genes for PC regression
genes <- select.gene(seurat.obj)$genes
#> Error: object 'seurat.obj' not found

# Scale data and perform PCA
seurat.obj <- prepare.seurat(seurat.obj, genes = genes)
#> Error: object 'genes' not found

# Build the KNN graph
seurat.obj <- prepare.graph(seurat.obj, knn = 30)
#> Error: object 'seurat.obj' not found

# Inspect the settings in NNet.setting
str(Seurat::Misc(seurat.obj, "NNet.setting"))
#> Error: object 'seurat.obj' not found