ceRNA Analysis

When to Use

Use this skill when you need to construct a ceRNA regulatory network from a known key-gene list using the bundled miRNA-mRNA and miRNA-lncRNA reference tables.

Use it for:

• Building a ceRNA network from one gene list and exporting flat CSV plus PDF outputs
• Comparing supported miRNA source modes such as combined, starbase, or pairwise overlaps
• Re-running the same local workflow with different lncRNA strictness, layout, or plotting parameters

Do not use it for:

• Differential expression, single-cell, enrichment, or survival analysis
• Workflows that do not start from a key gene list
• Cases where you want a miRNA-mRNA-only graph without a retained lncRNA ceRNA layer

Input Validation

This skill accepts:

• A key gene list as a plain-text file (one gene symbol per line) or as a comma-separated string on the CLI
• Optional parameter overrides for dataset mode, lncRNA strictness, layout, colors, and timeout

If the user's request does not involve building a ceRNA regulatory network from a key gene list — for example, asking to run differential expression, enrichment analysis, single-cell workflows, or survival analysis — do not proceed with the workflow. Instead respond:

"ceRNA Analysis is designed to construct a ceRNA regulatory network from a key gene list using bundled miRNA-mRNA and miRNA-lncRNA reference databases. Your request appears to be outside this scope. Please provide a key gene list and specify a supported miRNA dataset mode, or use a more appropriate skill for differential expression, enrichment analysis, or single-cell workflows."

When to Read External Files

Usage

Rscript scripts/main.R \
  --key_genes tests/data/gene.txt \
  --output_dir ./output/ \
  --mirna_dataset combined \
  --lncrna_strictness High \
  --lncrna_freq_thresh 0 \
  --timeout_seconds 600 \
  --seed 42

Dependency note: --help and all analysis modes require igraph to be installed. Install igraph before running any command. Use references/troubleshooting.md for installation guidance.

Arguments

Main Analysis: scripts/main.R

Input Format

Key Genes (key_genes)

Plain-text input with one gene symbol per line, or a comma-separated string passed directly on the CLI.

TP53
BRCA1
MYC

Rules:

• Blank lines are ignored
• Lines starting with # are ignored
• Duplicate genes are removed
• At least one valid gene is required

Database Directory (reference_dir)

The bundled database directory is references/database/. Required files depend on the selected mirna_dataset plus the selected lncRNA strictness file.

• combined: miRNA_mRNA.csv
• starbase: starbase_miRNA_mRNA.csv
• mirdb: miRDB_miRNA_mRNA.csv
• mirtarbase: miRTarbase_miRNA_mRNA.csv
• starbase+mirdb: starbase_miRNA_mRNA.csv and miRDB_miRNA_mRNA.csv
• starbase+mirtarbase: starbase_miRNA_mRNA.csv and miRTarbase_miRNA_mRNA.csv
• mirdb+mirtarbase: miRDB_miRNA_mRNA.csv and miRTarbase_miRNA_mRNA.csv
• lncRNA file: one of starbase_miRNA_lncRNA_High.csv, starbase_miRNA_lncRNA_Median.csv, or starbase_miRNA_lncRNA_Low.csv

Output Files

Workflow

Step 1: Validate Input

• Check key-gene input existence or parse comma-separated genes
• Validate parameter choices, numeric limits, timeout, and colors
• Verify the database directory and required files

Step 2: Load Interaction Data

• Load the selected miRNA-mRNA dataset
• Load the selected miRNA-lncRNA dataset by strictness level
• Recompute pairwise intersections when requested

Step 3: Filter the Network

• Retain miRNA-mRNA pairs linked to the provided key genes
• Retain miRNA-lncRNA pairs connected to the retained miRNAs
• Apply the lncRNA frequency threshold
• Stop with SKILL_INVALID_DATA if no lncRNA interactions remain after filtering, because the ceRNA layer has collapsed

Step 4: Build Outputs

• Construct edge and node tables
• Save CSV, PDF, and session information in the output directory root

Methods

combined

Uses the bundled precomputed overlap across three miRNA-mRNA resources for higher-confidence interactions.

Pairwise Intersections

starbase+mirdb, starbase+mirtarbase, and mirdb+mirtarbase recompute the overlap between two bundled databases. Pairwise intersections typically yield 20–40% fewer edges than combined mode because only interactions present in both selected databases are retained. Use pairwise modes when you need higher-confidence edges at the cost of reduced network coverage.

lncRNA Strictness

High, Median, and Low select different bundled starBase evidence levels for miRNA-lncRNA interactions.

Examples

Basic Combined Analysis

Rscript scripts/main.R \
  -i ./key_genes.txt \
  -o ./output \
  -m combined

Single Database Analysis

Rscript scripts/main.R \
  -i ./key_genes.txt \
  -o ./output_starbase \
  -m starbase \
  -l Median \
  -f 1

Error Handling

IF error persists, READ: references/troubleshooting.md

Testing

Test with Sample Data

# Run with sample data (igraph must be installed first)
Rscript scripts/main.R \
  -i tests/data/gene.txt \
  -o tests/output/

Validation Commands

# Inspect edge output
wc -l tests/output/ceRNA_network_edges.csv

# Check plot exists
ls -la tests/output/ceRNA_network.pdf