Rf Model Importance Analysis

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

RF Model Importance Analysis

Quick Start

Use one of these three commands first, then consult the full argument table only if you need extra tuning.

1. Standard Run

bashCopy code
Rscript scripts/main.R \
  --input_file tests/data/expression_matrix.csv \
  --group_file tests/data/group_info.csv \
  --case_group AR \
  --control_group Control \
  --output_dir tests/output/manual-test \
  --seed 42 \
  --timeout_seconds 300

2. Tuned Importance Run

bashCopy code
Rscript scripts/main.R \
  --input_file tests/data/expression_matrix.csv \
  --group_file tests/data/group_info.csv \
  --case_group AR \
  --control_group Control \
  --output_dir tests/output/custom-importance \
  --seed 42 \
  --rf_ntree 800 \
  --rf_mtry 4 \
  --rf_imp_type 2 \
  --rf_imp_threshold 1 \
  --rf_top_n 8 \
  --rf_importance_top_n 8 \
  --timeout_seconds 300

3. Plot-Only Rerender

Run this only after a full analysis has already created output_dir/data/rf_result.rds.

bashCopy code
Rscript scripts/main.R \
  --plot_only TRUE \
  --output_dir tests/output/manual-test \
  --seed 42 \
  --timeout_seconds 300

When to Read External Files

Stop Conditions

Do not use this skill when any of the following is true:

• The task is regression, multiclass classification, time-series modeling, or remote data fetching.
• The input still requires imputation, normalization, batch correction, or other preprocessing.
• The feature matrix contains missing values, non-numeric feature columns, or mismatched sample IDs.

If one of those conditions applies, stop and hand off to a preprocessing or alternative modeling workflow before running this skill.

Usage

Before running the CLI, ensure the data is already cleaned for binary classification: samples in rows, numeric feature columns only, and no missing values. Imputation, normalization, and batch correction are outside this skill's scope.

bashCopy code
Rscript scripts/main.R \
  --input_file ./input/expression_matrix.csv \
  --group_file ./input/group_info.csv \
  --case_group Case \
  --control_group Control \
  --output_dir output/basic-run \
  --seed 42 \
  --timeout_seconds 600

Arguments

Input Format

Expression Matrix

• CSV or TSV.
• First column: sample IDs.
• Remaining columns: numeric features.
• Samples must be rows.
• Missing or non-numeric feature values are not allowed.

Example:

csvCopy code
sample,HIF1A,NR4A1,SOCS1
S1,6.21,-1.34,2.01
S2,6.57,0.37,3.62
S3,7.05,2.12,5.01

Group File

• CSV or TSV.
• First column: sample IDs.
• One additional column must contain both the case and control labels.
• Exactly two groups are supported.

Example:

csvCopy code
sample,group
S1,Case
S2,Case
S3,Control

Output Files

Error Handling

• Successful runs exit with status code 0.
• Failed runs exit with status code 1.
• Error messages use standardized names such as SKILL_FILE_NOT_FOUND and SKILL_INVALID_PARAMETER.
• Output paths are validated so that --output_dir cannot write outside the skill root.
• The analysis never performs network requests and never executes user input through eval(), exec(), or system().

Common codes:

For detailed fixes, READ: references/troubleshooting.md

Testing

Help Check

bashCopy code
Rscript scripts/main.R --help

Full Test Run

bashCopy code
Rscript tests/run_tests.R

Direct Test Command

bashCopy code
Rscript scripts/main.R \
  --input_file tests/data/expression_matrix.csv \
  --group_file tests/data/group_info.csv \
  --case_group AR \
  --control_group Control \
  --output_dir tests/output/manual-test \
  --seed 42 \
  --rf_ntree 200 \
  --rf_top_n 5 \
  --rf_importance_top_n 5 \
  --timeout_seconds 300