LoadingRNAFromSeurat Process Configuration

Purpose

Load pre-existing Seurat objects into the immunopipe pipeline instead of starting from raw count matrices via SampleInfo. This enables analysis on already processed single-cell RNA-seq data stored in Seurat R objects.

When to Use

• Starting from pre-processed Seurat objects (RDS or qs/qs2 format) instead of raw count matrices
• Re-analyzing existing Seurat objects with immunopipe's downstream analysis capabilities
• Alternative entry point when SampleInfo is not needed for RNA data input
• When metadata is already embedded in the Seurat object's meta.data slot
• When combining with TCR/BCR data - can use LoadingRNAFromSeurat for RNA + SampleInfo for VDJ data

Configuration Structure

Process Enablement

[LoadingRNAFromSeurat]
cache = true

Input Specification

[LoadingRNAFromSeurat.in]
# Path to Seurat object file (RDS or qs/qs2 format)
# Can be single file or array of files for multiple samples
infile = ["path/to/seurat_object.rds"]

# Alternative: can use 'srtobj' alias (same as infile)
# srtobj = ["path/to/seurat_object.rds"]

Environment Variables

[LoadingRNAFromSeurat.envs]
# Whether the Seurat object is well-prepared for the pipeline
# - If true: SeuratPreparing process will be skipped
# - If false: SeuratPreparing will run for QC, normalization, integration
prepared = false

# Whether the Seurat object is already clustered
# - If true: SeuratClustering (or SeuratClusteringOfAllCells) and SeuratMap2Ref will be skipped
# - Forces 'prepared' to be true if set to true
clustered = false

# Column name in Seurat object's meta.data that contains sample identifiers
# Used to create a "Sample" column in the output
# Default is "Sample" - if meta.data already has "Sample", no action is taken
# If column exists but named differently, specify here (e.g., "orig.ident", "sample_id")
sample = "Sample"

Configuration Examples

Minimal Configuration (Single Seurat Object)

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["path/to/sample1.rds"]

Pre-processed Seurat Object (Skip Preparation)

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/preprocessed_seurat.rds"]

[LoadingRNAFromSeurat.envs]
# Object already normalized, QC'd, integrated - skip SeuratPreparing
prepared = true

Fully Prepared Object with Clustering

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/clustered_seurat.rds"]

[LoadingRNAFromSeurat.envs]
# Object is fully prepared and clustered
# Skip both SeuratPreparing and SeuratClustering
clustered = true
# 'prepared' automatically set to true when clustered = true

Custom Sample Column Mapping

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/seurat_objects/sample1.rds", "data/seurat_objects/sample2.rds"]

[LoadingRNAFromSeurat.envs]
# Seurat object uses "orig.ident" column for sample names
sample = "orig.ident"

Loading Multiple Seurat Objects

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = [
    "data/sample1.rds",
    "data/sample2.rds",
    "data/sample3.rds"
]

[LoadingRNAFromSeurat.envs]
# Each object must have the sample column specified
# Objects will be integrated by SeuratPreparing if prepared = false
sample = "Sample"

RNA + TCR Combined Analysis

# Use LoadingRNAFromSeurat for RNA data
[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/rna_seurat.rds"]
[LoadingRNAFromSeurat.envs]
prepared = true

# Still use SampleInfo for TCR/BCR data paths
[SampleInfo.in]
infile = ["sample_info.txt"]
# sample_info.txt should contain TCRData/BCRData columns (not RNAData)

Common Patterns

Pattern 1: Load and Start Analysis (Standard Workflow)

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/seurat.rds"]

# SeuratPreparing will run for QC, normalization, integration
# SeuratClustering will run for clustering
[SeuratClustering]
[SeuratClusterStats]

Pattern 2: Load and Skip to Downstream Analysis

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/prepared_seurat.rds"]
[LoadingRNAFromSeurat.envs]
prepared = true  # Skip SeuratPreparing

# Jump directly to clustering and marker analysis
[SeuratClustering]
[ClusterMarkers]
[SeuratClusterStats]

Pattern 3: Fully Pre-processed (Skip Preparation + Clustering)

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/final_seurat.rds"]
[LoadingRNAFromSeurat.envs]
clustered = true  # Skip SeuratPreparing AND SeuratClustering

# Jump directly to downstream analyses
[CellTypeAnnotation]
[ScFGSEA]

Pattern 4: TCR Analysis with Pre-processed RNA

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = ["data/rna_seurat.rds"]
[LoadingRNAFromSeurat.envs]
prepared = true

# Still load TCR/BCR data
[ScRepLoading]

# Continue with TCR-specific analyses
[TOrBCellSelection]
[CDR3Clustering]
[ClonalStats]

Pattern 5: Multi-sample Integration

[LoadingRNAFromSeurat]
[LoadingRNAFromSeurat.in]
infile = [
    "data/patient1.rds",
    "data/patient2.rds",
    "data/patient3.rds"
]
[LoadingRNAFromSeurat.envs]
# Each object has a "patient_id" column for sample identification
sample = "patient_id"

# SeuratPreparing will integrate multiple samples
[SeuratPreparing]
[SeuratClustering]

Dependencies

Upstream

• None (Entry point process)
• Can optionally work with SampleInfo when TCR/BCR data is present (SampleInfo provides VDJ paths)

Downstream

• SeuratPreparing (if prepared = false)
  • Performs QC, normalization, integration of loaded Seurat objects
  • Required for standard analysis workflow
• SeuratClustering or SeuratClusteringOfAllCells (if clustered = false)
  • Performs clustering analysis
• All downstream RNA analysis processes:
  • SeuratClusterStats, ClusterMarkers, CellTypeAnnotation, SeuratMap2Ref, etc.

Validation Rules

File Format Requirements

• Supported formats: RDS (saveRDS() / readRDS()) or qs/qs2 (qs::qsave() / qs::qread())
• Content: Must contain a valid Seurat object
• File existence: Input files must exist at specified paths
• Sample column: If sample parameter is not "Sample", the specified column must exist in object@meta.data

Metadata Handling

• If meta.data already contains a "Sample" column and sample = "Sample":
  • No modification is made (symlink created to save space)
• If sample column doesn't exist:
  • Error: Process fails with message "Sample column 'X' not found in metadata"
• If sample column exists with custom name (not "Sample"):
  • A new "Sample" column is created by copying from the specified column
  • Modified object is saved to output

SampleInfo Compatibility

• Mutually exclusive with RNAData: Cannot use both LoadingRNAFromSeurat and RNAData column in SampleInfo
• Compatible with TCRData/BCRData: Can use LoadingRNAFromSeurat for RNA + SampleInfo for VDJ data paths
• Required when: No SampleInfo section exists AND RNA data is needed

Environment Variable Validation

• clustered = true → automatically sets prepared = true (forced dependency)
• sample column must exist in Seurat object metadata
• Boolean flags accept true/false (case-insensitive in TOML)

Troubleshooting

Issue: "Sample column not found in metadata"

Cause: The specified sample column name doesn't exist in object@meta.data Solution:

[LoadingRNAFromSeurat.envs]
# Check your Seurat object's metadata:
# colnames(seurat_obj@meta.data)
sample = "actual_column_name"  # Use the exact column name

Issue: SeuratPreparing still running despite prepared = true

Cause: Configuration syntax error or caching issue Solution:

1. Check TOML syntax (no quotes around boolean values)
2. Clear cache: [LoadingRNAFromSeurat] cache = "force"
3. Verify config is being loaded: python -m immunopipe.validate_config config.toml

Issue: Multiple samples not being integrated

Cause: Sample column mapping incorrect or objects don't have the specified column Solution:

[LoadingRNAFromSeurat.in]
infile = ["sample1.rds", "sample2.rds"]
[LoadingRNAFromSeurat.envs]
# Verify each object has this column before running
sample = "orig.ident"  # Common alternative to "Sample"

Issue: Want to use LoadingRNAFromSeurat but also have TCR data

Cause: Unclear how to specify TCR data paths Solution: Use both processes:

[LoadingRNAFromSeurat.in]
infile = ["rna_seurat.rds"]

[SampleInfo.in]
infile = ["sample_info.txt"]
# sample_info.txt only needs TCRData/BCRData columns (not RNAData)

Issue: Symlink error when Sample column already exists

Cause: Trying to create symlink when file exists Solution: This is handled automatically by the script - it removes existing file before creating symlink

Issue: Want to combine LoadingRNAFromSeurat with SampleInfo metadata

Cause: Need additional metadata columns Solution: Use SeuratPreparing.envs.mutaters to add columns:

[SeuratPreparing.envs]
mutaters = {
    "Condition" = "metadata$Condition",
    "Batch" = "metadata$Batch"
}

Best Practices

1. Always specify sample column: Even if default is "Sample", explicitly set it to avoid issues
2. Check metadata before running: Use R to verify column names exist in object@meta.data
3. Use prepared = true for re-analysis: Skip unnecessary preprocessing when objects are already prepared
4. Use clustered = true cautiously: Only skip clustering if you're satisfied with existing clustering
5. Validate configuration: Run python -m immunopipe.validate_config config.toml before executing pipeline
6. Consider file size: Large RDS files can be slow to copy; use qs/qs2 format for better performance

Difference from SampleInfo

Feature	SampleInfo	LoadingRNAFromSeurat
Input format	Raw count matrices (10X, loom)	Pre-processed Seurat objects
Data preparation	Always requires SeuratPreparing	Optional (can skip with prepared = true)
Metadata source	Sample info text file	Embedded in Seurat object
Multi-sample handling	Specified in text file	Multiple input files or single multi-sample object
TCR/BCR data support	Provides paths for RNA + VDJ	Only RNA (use SampleInfo for VDJ)
Integration	Required step	Depends on prepared setting

Workflow Integration

LoadingRNAFromSeurat replaces the standard SampleInfo → SeuratPreparing entry point:

Standard workflow (raw data):

SampleInfo → SeuratPreparing → SeuratClustering → downstream analyses

With LoadingRNAFromSeurat (prepared data):

LoadingRNAFromSeurat → SeuratClustering → downstream analyses

With LoadingRNAFromSeurat (fully processed):

LoadingRNAFromSeurat → downstream analyses (skip SeuratClustering)

With TCR data:

LoadingRNAFromSeurat (RNA) + SampleInfo (VDJ paths) → ScRepLoading → TCR analyses