Weka format pathway export

weka-format-pathway-export

Summary

Export information-gain ranked pathways from a pathway dysregulation score (PDS) matrix into Weka-compatible format for downstream machine learning classification. This skill bridges personalized pathway analysis with standardized feature selection output suitable for phenotype prediction.

When to use

You have computed a pathway dysregulation score matrix (PDSmatrix) from metabolomics data and need to rank pathways by their relevance to a binary or multi-class phenotype label using information-gain scoring, then export the ranked set for use in machine learning classifiers (SVM, Random Forest, LDA, etc.). Apply this when your input is metabolite-pathway associations and phenotype labels, and your goal is phenotype prediction via pathway features.

When NOT to use

• PDSmatrix has not yet been computed from metabolite-pathway associations (use lilikoi.PDSfun first).
• Phenotype labels are continuous (e.g., survival time, biomarker concentration) rather than categorical; use regression-based feature selection instead.
• You are performing unsupervised pathway discovery with no phenotype annotation; information-gain requires labeled data.

Inputs

• PDSmatrix (numeric matrix: pathways × samples with dysregulation scores)
• phenotype labels (factor or character vector: class labels per sample)
• information-gain threshold (numeric, typical range 0.30–0.70)

Outputs

• selected_Pathways_Weka (data frame or text file: ranked pathway names with information-gain scores)
• information-gain scores per pathway (numeric vector)

How to apply

Load the PDSmatrix (pathway-by-sample feature matrix) and corresponding phenotype labels into R. Apply the lilikoi.featuresSelection function with method='gain' (information-gain) and a threshold (typically 0.50) to rank pathways by their mutual information with the phenotype and filter to only those exceeding the threshold. The function returns selected_Pathways_Weka, a ranked list of pathway names and their information-gain scores in a format compatible with Weka machine learning workflows. The threshold acts as a specificity control: higher thresholds (e.g., 0.70) retain only the most discriminative pathways, while lower thresholds (e.g., 0.30) retain more candidate pathways at the risk of including noise. The information-gain metric quantifies how much knowing a pathway's activity reduces uncertainty about the phenotype class.

Related tools

• Lilikoi v2.0 (R package hosting lilikoi.featuresSelection function for information-gain pathway ranking and Weka export) — https://github.com/lanagarmire/lilikoi2
• R (Programming environment for executing lilikoi.featuresSelection and managing PDSmatrix objects)
• Weka (Target machine learning framework for which the selected_Pathways_Weka output is formatted)

Examples

selected_Pathways_Weka <- lilikoi.featuresSelection(PDSmatrix, threshold=0.50, method="gain")

Evaluation signals

• selected_Pathways_Weka contains only pathways with information-gain scores ≥ threshold; verify no scores fall below the specified cutoff.
• Pathway list is ranked in descending order by information-gain score; top pathways are more discriminative for the phenotype.
• Output file format is compatible with Weka (tab- or comma-delimited; pathway name and score columns are present and parseable).
• Number of selected pathways is substantially smaller than input PDSmatrix row count, indicating effective filtering (typical: 5–50 pathways selected from thousands).
• Information-gain scores are bounded in [0, log₂(number of classes)]; for binary phenotypes, max score ≤ 1.0.

Limitations

• Information-gain assumes categorical phenotype; continuous outcomes require alternative feature selection methods (e.g., correlation, regression coefficient ranking).
• Threshold selection (0.50 in the reference workflow) is heuristic; no data-driven method for optimizing it is described; requires domain knowledge or cross-validation tuning.
• Pathways with rare or zero information-gain scores in small sample sets may be filtered out, potentially missing rare but clinically relevant signals.
• No correction for multiple comparisons applied; information-gain scoring does not account for multiple testing burden across pathways.

Evidence

• [other] Apply information-gain method for feature selection: "Apply the information-gain method for feature selection with threshold 0.50 in Lilikoi v2.0 to rank and filter pathways by their relevance to the phenotype."
• [other] PDSmatrix is the pathway dysregulation score matrix: "Load the PDSmatrix (pathway-by-sample feature matrix) and phenotype labels into R."
• [other] Output is selected_Pathways_Weka: "producing selected_Pathways_Weka as output."
• [other] Information-gain scores rank pathway relevance: "to rank and filter pathways by their relevance to the phenotype."
• [readme] Lilikoi.featuresSelection implementation in readme: "selected_Pathways_Weka= lilikoi.featuresSelection(PDSmatrix,threshold= 0.50,method="gain")"