SeuratClusteringOfAllCells Process Configuration

Purpose

Performs coarse clustering on ALL cells (including T cells, B cells, and non-T/B cells) before cell type selection. This process identifies broad cell populations to enable subsequent T/B cell selection via TOrBCellSelection. Unlike SeuratClustering which works on already-selected T/B cells, this provides initial clustering on heterogeneous cell populations.

When to Use

• Mixed cell populations: When your data contains both T/B cells AND non-T/B cells
• Pre-selection clustering: Required upstream of TOrBCellSelection process
• Broad cell type identification: To identify major cell lineages before fine-grained analysis
• TCR/BCR data analysis: When you have scRNA-seq + scTCR/scBCR data with mixed populations
• Do NOT use when: All cells are already T/B cells (use SeuratClustering instead)

Configuration Structure

Process Enablement

[SeuratClusteringOfAllCells]
cache = true

Input Specification

[SeuratClusteringOfAllCells.in]
srtobj = ["SeuratPreparing"]

Environment Variables

Core Parameters

[SeuratClusteringOfAllCells.envs]
ncores = 1
ident = "seurat_clusters"
cache = "/tmp"

FindNeighbors Parameters

[SeuratClusteringOfAllCells.envs.FindNeighbors]
k.param = 20
reduction = "pca"
dims = 30
prune.SNN = 0.067

RunUMAP Parameters

[SeuratClusteringOfAllCells.envs.RunUMAP]
reduction = "pca"
dims = 30
n.neighbors = 30
min.dist = 0.3
seed.use = 42

FindClusters Parameters

[SeuratClusteringOfAllCells.envs.FindClusters]
resolution = 0.5  # Use LOWER (0.2-0.8) for coarse clustering
algorithm = 4  # 4 = Leiden (recommended)
random.seed = 0
graph.name = "pca_snn"

External References

• FindNeighbors: https://satijalab.org/seurat/reference/findneighbors
• RunUMAP: https://satijalab.org/seurat/reference/runumap
• FindClusters: https://satijalab.org/seurat/reference/findclusters

All parameters identical to SeuratClustering.

Configuration Examples

Minimal Configuration

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.in]
srtobj = ["SeuratPreparing"]

Standard Pre-selection Clustering

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.envs.FindClusters]
resolution = 0.4
algorithm = 4

Multiple Resolutions

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.envs.FindNeighbors]
k.param = 25

[SeuratClusteringOfAllCells.envs.FindClusters]
resolution = [0.2, 0.4, 0.6]
algorithm = 4

Integrated Data

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.envs.FindNeighbors]
reduction = "integrated.cca"

[SeuratClusteringOfAllCells.envs.RunUMAP]
reduction = "integrated.cca"

[SeuratClusteringOfAllCells.envs.FindClusters]
resolution = 0.5

Common Patterns

Pattern 1: Coarse Clustering for Cell Type ID

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.envs.FindClusters]
resolution = 0.3
algorithm = 4

Pattern 2: Resolution Sweep

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.envs.FindClusters]
resolution = "0.2:0.8:0.2"
algorithm = 4

Pattern 3: Large Datasets

[SeuratClusteringOfAllCells]
[SeuratClusteringOfAllCells.envs]
ncores = 8

[SeuratClusteringOfAllCells.envs.FindNeighbors]
nn.method = "annoy"
dims = 25

Dependencies

Upstream

• Required: SeuratPreparing

Downstream

• Required: TOrBCellSelection
• Optional: ClusterMarkersOfAllCells, TopExpressingGenesOfAllCells

Validation Rules

Resolution Constraints

• Must be positive, single value or list
• Recommendation: Use lower resolutions (0.2-0.8)

Algorithm Selection

• Leiden (algorithm=4) recommended

Troubleshooting

Issue: T/B Cells Not Separated

Solution: Lower resolution to 0.3, increase k.param to 30

Issue: Too Many Clusters

Solution: Use coarse resolution (0.2)

Issue: Poor UMAP Separation

Solution: min.dist = 0.1, n.neighbors = 15

Key Differences from SeuratClustering

Feature	SeuratClusteringOfAllCells	SeuratClustering
Timing	BEFORE T/B selection	AFTER T/B selection
Data scope	ALL cells (mixed)	Selected T/B cells
Resolution	LOWER (0.2-0.8)	HIGHER (0.8-1.5)
Purpose	Identify major lineages	Sub-cluster T/B

Best Practices

1. Use lower resolutions (0.2-0.8)
2. Follow with TOrBCellSelection
3. Leiden algorithm (algorithm=4) recommended
4. Set random seeds for reproducibility
5. Don't use when all cells are T/B cells

Related Processes

• TOrBCellSelection: Selects T/B cells
• SeuratClustering: Fine-grained clustering
• ClusterMarkersOfAllCells: Marker analysis before selection