Single Cell Data Export Pipeline

Overview

This R script exports Seurat single-cell RNA-seq data to tab-separated format for downstream analysis. It processes large datasets using chunked memory-efficient operations.

Output Files

• Dense matrix: *_gene_by_cell_matrix.tsv - Log-normalized expression values (genes × cells)
• UMAP coordinates: umap-*.tsv - 2D embedding coordinates
• Cell metadata: metadata-*.tsv - Cell annotations and QC metrics

Resource Requirements

Memory Allocation Guidelines

Dataset Size	Memory	CPUs	Time Estimate
~300K cells	800GB	12	~15 minutes
~150K cells	400GB	8	~8 minutes
~75K cells	200GB	6	~4 minutes
~30K cells	100GB	4	~2 minutes

Galyleo Launch Command Template

galyleo launch -p platinum -q hcp-csd854 -A csd854 \
  --time-limit 9:00:00 \
  --conda-env r44_env_v2 \
  --memory [MEMORY_GB] \
  --cpus [CPU_COUNT]

Example for Different Dataset Sizes

# Large dataset (300K+ cells)
galyleo launch -p platinum -q hcp-csd854 -A csd854 --time-limit 9:00:00 --conda-env r44_env_v2 --memory 800 --cpus 12

# Medium dataset (100-300K cells)
galyleo launch -p platinum -q hcp-csd854 -A csd854 --time-limit 6:00:00 --conda-env r44_env_v2 --memory 400 --cpus 8

# Small dataset (<100K cells)
galyleo launch -p platinum -q hcp-csd854 -A csd854 --time-limit 3:00:00 --conda-env r44_env_v2 --memory 200 --cpus 6

Usage Notes

• Adjust chunk_size parameter if memory issues persist
• Monitor peak memory usage stays under allocated limits
• Dense matrix files can be very large (>100GB for large datasets)
• Processing time scales roughly linearly with cell count