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

Kaplan-Meier Survival Curve Analysis

Use this skill to run Kaplan-Meier survival analysis on a tabular dataset and export a single PDF survival figure.

Use This Skill When

• You need a Kaplan-Meier survival curve from a table containing time, status, and group columns.
• You need a command-line survival workflow with parameter validation.
• You need a single Kaplan-Meier plot as the final analysis result.

Primary Command

Rscript scripts/main.R \
  --input_file <input_file> \
  --output_dir <output_dir> \
  --time_col <time_column> \
  --status_col <status_column> \
  --risk_col <group_column>

Prerequisites

• Rscript is available in the shell.
• Required R packages: optparse, data.table, survival, survminer, ggplot2.
• Install missing packages with Rscript -e 'install.packages(c("optparse", "data.table", "survival", "survminer", "ggplot2"), repos="https://cloud.r-project.org")'.

Core Arguments

Argument	Required	Description
--input_file	Yes	Input data file in CSV or tab-delimited TXT/TSV format
--output_dir	No	Output directory, default ./KM_Results
--time_col	No	Survival time column, default futime
--status_col	No	Event status column, default fustat
--risk_col	No	Risk group column, default risk_group
--time_unit	No	Time unit label: year, month, or day, default year
--auto_convert_days	No	Heuristically convert large time values from days when time_unit is year or month, default true
--statistics_method	No	logrank or wald, default logrank

Plot Customization Arguments

Argument	Default	Description
--figure_width	10	Figure width in inches
--figure_height	7	Figure height in inches
--figure_family	sans	Font family
--title_x	Time	X-axis title. If left as default, the script renders Time (<time_unit>)
--title_y	Survival probability	Y-axis title
--title_main	empty	Plot title
--legend_position	top	top, bottom, left, right, none
--legend_show	true	Whether to show legend
--legend_title	empty	Legend title
--line_type	solid	Survival line type: solid, dashed, dotted, dotdash, longdash, twodash
--line_size	1	Survival line width
--line_colors	#4DBBD5,#E64B35,#00A087,#3C5488,#F39B7F,#8491B4,#91D1C2,#DC0000	Comma-separated group colors
--censor_show	true	Whether to show censor markers
--censor_size	7	Censor marker size
--confidence_show	true	Whether to show confidence interval
--confidence_alpha	0.2	Confidence band transparency
--risk_table_show	true	Whether to show the risk table
--risk_table_border	true	Whether to show the risk table border
--risk_table_panel	false	Whether to show the risk table panel background
--risk_table_size	6	Risk table font size
--axis_title_size	12	Axis title font size
--axis_text_size	10	Axis tick-label font size
--legend_text_size	11	Legend text font size

Input Requirements

• The input file must contain the requested time, status, and group columns.
• .txt inputs must be tab-delimited.
• time must contain finite non-negative numeric values.
• status must be coded as 0 for censored and 1 for event.
• The risk group column must be a precomputed categorical grouping variable, not a continuous score column.
• The risk group column must contain at least 2 groups after filtering.
• --line_colors must provide at least one color per retained group when you override the default palette.
• Rows with missing time, status, or group values are removed before analysis.
• At least 2 complete observations must remain after filtering.
• Near-unique or continuous-looking grouping columns are rejected before model fitting.
• If --auto_convert_days true and max(time) > 365, the script assumes the retained time values are in days and converts them to the requested --time_unit when time_unit is year or month.
• Only use --auto_convert_days true when the source time column is known to be in days.
• If your source data are already in years or months, disable --auto_convert_days to avoid incorrect conversion.
• --statistics_method wald only supports exactly 2 retained groups; use logrank for multi-group comparisons.
• Invalid plotting parameters such as unsupported --line_type values are rejected before plotting.

Example input:

id	fustat	futime	risk_score	risk_group	GPR161	RIBC2
TCGA-C5-A1M5	1	5.62191780821918	-1.10702407761445	low	2.82521576230566	5.35318564979635
TCGA-VS-A94W	0	3.40547945205479	-0.671246677921865	high	4.26241812321536	4.00802068790173

Bundled test datasets:

• tests/data/km_sample1.txt: baseline KM example with risk_group
• tests/data/km_sample2.txt: alternate cohort for plotting and statistics examples
• tests/data/km_sample3.txt: additional cohort for validation and repeated testing

Minimal Workflow

1. Confirm the input file exists and identify the time, status, and group columns.
2. Run scripts/main.R with the requested output directory and any optional plot parameters.
3. Check the output directory for km-plot.pdf.

If you omit --input_file, the script exits with SKILL_MISSING_INPUT.

Outputs

Expected output:

<output_dir>/
├── km-plot.pdf
└── session_info.txt

Interpretation Guide

• Use the survival figure to inspect separation between groups over time.
• Use the p-value annotation, confidence interval, and risk table in the figure to interpret group separation.

Time Conversion Caution

• Automatic conversion is a convenience heuristic, not a unit detector.
• The script only checks whether max(time) > 365; it does not infer the true source unit from metadata.
• If the input time column is already expressed in years or months, run with --auto_convert_days false.
• Review the console log for the conversion warning whenever time_unit is year or month.

Reproducibility Note

• Repeated runs on identical input should be analytically consistent.
• The exported km-plot.pdf may not be byte-identical across repeated runs because PDF metadata and graphics-device output can vary.
• If you need byte-stable artifacts, add a deterministic PDF post-processing step outside this skill.

Do Not Use This Skill When

• You need this tool to derive a cutoff or split a continuous score into risk groups.
• You need multivariable Cox regression, covariate adjustment, or hazard-ratio modeling beyond the p-value route already exposed here.
• You need a broader survival-analysis workflow with upstream feature engineering, biomarker selection, or data harmonization.
• You need multiple plots, report generation, or downstream interpretation beyond producing one Kaplan-Meier figure and session metadata.

Read These Files When Needed

Need	File
Kaplan-Meier method details and interpretation	references/algorithm.md
More CLI examples	references/cli-guide.md
Error diagnosis	references/troubleshooting.md
Main execution entry point	scripts/main.R
Sample test data	tests/data/km_sample1.txt, tests/data/km_sample2.txt, tests/data/km_sample3.txt

Quick Examples

Basic Kaplan-Meier analysis:

Rscript scripts/main.R \
  --input_file tests/data/km_sample1.txt \
  --output_dir tests/output_basic

Custom column names:

Rscript scripts/main.R \
  --input_file tests/data/km_sample1.txt \
  --time_col futime \
  --status_col fustat \
  --risk_col risk_group \
  --output_dir tests/output_custom_columns

Custom plot title:

Rscript scripts/main.R \
  --input_file tests/data/km_sample2.txt \
  --title_main "Study KM Curve" \
  --output_dir tests/output_title

Hide confidence interval and risk table:

Rscript scripts/main.R \
  --input_file tests/data/km_sample3.txt \
  --confidence_show false \
  --risk_table_show false \
  --output_dir tests/output_simple

Validation

Rscript scripts/main.R --help

Rscript scripts/main.R \
  --input_file tests/data/km_sample1.txt \
  --output_dir tests/validation_output

After running analysis, verify that tests/validation_output/km-plot.pdf exists.

Common Errors

• SKILL_FILE_NOT_FOUND: Input file path is wrong or inaccessible.
• SKILL_MISSING_COLUMNS: A requested time, status, or risk group column is missing.
• SKILL_INVALID_DATA: Input data is malformed or unsuitable for survival analysis.
• SKILL_INVALID_DATA: A continuous or near-unique risk column was supplied where a categorical group column is required.
• SKILL_INVALID_PARAMETER: An argument value is invalid.
• SKILL_INVALID_PARAMETER: Plotting options such as --line_type or --line_colors are incompatible with the retained groups.
• SKILL_INSUFFICIENT_DATA: Too few complete observations remain after filtering.
• SKILL_DEPENDENCY_MISSING: A required R package such as optparse or survival is unavailable.

If the issue is not obvious, read references/troubleshooting.md.