Source: https://github.com/aipoch/medical-research-skills

Consensus Clustering Analysis

When to Use

Use this skill when you need to identify stable sample subtypes from a bulk expression matrix with ConsensusClusterPlus, compare candidate clustering settings with PAC, and export consensus matrix/CDF visualizations.

Do not use this skill for differential expression analysis, single-cell clustering, or non-expression tabular data.

When to Read External Files

Situation	File to Read	Purpose
Need algorithm details	references/algorithm.md	Consensus clustering, PAC scoring, and preprocessing assumptions
Need to run analysis	scripts/main.R	Execute: Rscript scripts/main.R --input_file ... --group_file ...
Encounter errors	references/troubleshooting.md	Common errors and solutions
Need CLI examples	references/cli-guide.md	Detailed CLI usage examples with verified local runs

---

Usage

Rscript scripts/main.R \
  --input_file ./expression_matrix.csv \
  --group_file ./groups.csv \
  --disease_group case \
  --max_k 4 \
  --output_dir ./output/ \
  --gene_selection highly_variable \
  --top_n 5000 \
  --reps 1000 \
  --p_item 0.8 \
  --p_feature 1.0 \
  --timeout_seconds 3600 \
  --seed 42

---

Arguments

Short	Long	Type	Default	Description
-i	--input_file	character	required	Expression matrix file (genes as rows, samples as columns)
-g	--group_file	character	required	Group information file (sample ID + group columns)
-d	--disease_group	character	case	Group label retained for clustering
-k	--max_k	integer	4	Maximum cluster count to evaluate
-o	--output_dir	character	./output/	Output directory
-m	--gene_selection	character	highly_variable	Gene selection mode: highly_variable or custom
-n	--top_n	integer	5000	Number of top variable genes to keep
-l	--gene_list	character	NULL	Custom gene list file when gene_selection=custom
-c	--center_data	logical	TRUE	Median-center each gene before clustering
-r	--reps	integer	1000	Consensus resampling repetitions
	--p_item	double	0.8	Sample resampling proportion
	--p_feature	double	1.0	Feature resampling proportion
-t	--timeout_seconds	integer	3600	Elapsed timeout in seconds
-s	--seed	integer	42	Random seed for reproducibility

---

Input Format

Expression Matrix (input_file)

Genes as rows, samples as columns, CSV/TSV/TXT format with gene ID in the first column.

,Sample01,Sample02,Sample03
TSPAN6,1.8479,1.8318,3.8276
TNMD,0.0349,0.0533,1.3889

Group File (group_file)

Delimited text file with sample ID and group columns.

sample,group
Sample01,case
Sample02,control
Sample03,case

Gene List (gene_list)

Optional plain text or single-column CSV file with one gene symbol per line.

TNMD
DPM1
SCYL3

---

Output Files

File	Description
Cluster_res.csv	PAC summary for each distance/algorithm combination with is_best marking the selected model
genes_for_clustering.csv	Selected genes and gene selection mode
samples_for_clustering.csv	Samples retained after disease-group filtering
result_<distance>_<algorithm>/	Method-specific consensus outputs and PAC_scores.csv
Consensus Matrix Plot.pdf	Consensus matrix heatmap for the optimal model
CDF curve Plot.pdf	CDF curves for the optimal method
session_info.txt	R session and package version info

---

Workflow

Step 1: Validate Input

• Check file existence
• Detect sample and group columns in the group file
• Validate sample matching between expression matrix and group file

Step 2: Prepare Clustering Matrix

• Filter samples by the requested disease group
• Select genes using highly_variable or custom
• Median-center genes if requested

Step 3: Run Consensus Clustering

• Evaluate supported distance and clustering algorithm combinations
• Compute PAC scores across candidate K values
• Select the optimal model by minimum PAC

Step 4: Generate Outputs

• Save result tables
• Generate consensus matrix and CDF plots
• Record session information for reproducibility

---

Methods

ConsensusClusterPlus

Repeated subsampling is used to estimate cluster stability across candidate K values and clustering settings.

PAC Score

The proportion of ambiguous clustering is computed as CDF(0.9) - CDF(0.1) from lower-triangle consensus values. Lower PAC indicates more stable clustering.

Gene Selection

• highly_variable: rank genes by median absolute deviation
• custom: use the intersection of the provided gene list and matrix row names

---

Examples

Basic Usage

Rscript scripts/main.R \
  -i expression_matrix.csv \
  -g groups.csv \
  -d case \
  -k 3 \
  -r 20 \
  -o output/example_basic \
  -t 120

With a Custom Gene List

Rscript scripts/main.R \
  -i expression_matrix.csv \
  -g groups.csv \
  -d case \
  -m custom \
  -l genes.csv \
  -k 4 \
  -r 20 \
  -o output/example_custom \
  -t 120

Without Median Centering

Rscript scripts/main.R \
  -i expression_matrix.csv \
  -g groups.csv \
  -d case \
  -c FALSE \
  -k 3 \
  -r 20 \
  -o output/example_rawscale \
  -t 120

---

Error Handling

Common Errors

Error	Cause	Solution
SKILL_FILE_NOT_FOUND	Input file does not exist	Check file path and permissions
SKILL_MISSING_COLUMNS	Group file lacks sample/group columns	Verify column names in the group file
SKILL_SAMPLE_MISMATCH	Sample names do not match	Ensure group file sample IDs match matrix columns
SKILL_INVALID_PARAMETER	CLI value is invalid	Check allowed options and numeric ranges
SKILL_INVALID_DATA	Too few samples/genes remain after filtering	Lower max_k or review the input data
SKILL_TIMEOUT	Run exceeded the configured timeout	Increase timeout_seconds or reduce reps
SKILL_DEPENDENCY_MISSING	Required R package is not installed	Install missing packages before rerunning

IF error persists, READ: references/troubleshooting.md

---

Testing

Smoke Check

# Check help
Rscript scripts/main.R --help

# Run analysis
Rscript scripts/main.R \
  -i tests/data/expression_matrix.csv \
  -g tests/data/groups.csv \
  -d case \
  -k 3 \
  -r 20 \
  -o output/example_basic \
  -t 120

Validation Commands

# Inspect selected model
cat output/example_basic/Cluster_res.csv

# Check output plots exist
ls -la output/example_basic

---

Implementation Checklist

• CLI parsing with optparse
• set.seed() for reproducibility
• requireNamespace() dependency checks
• Session info recording
• data.table::fread() input reading
• File reading instructions in SKILL.md
• Modular script structure (<150 lines per file)
• Test data provided
• Error handling with SKILL_* codes
• Scripts in scripts/ directory
• References in references/ directory

---

Last updated: 2026-04-17 | Version: 1.0.0