Skip to contents

Changelog

Source: NEWS.md

WSSmod 0.1.0

Initial versioned release.

Module scores

  • calculate_WSS() computes Weighted Sum of Scores module-level summaries from proteomic data, using STRING network connectivity and biomarker-association weighting to align and aggregate protein-level signal into module scores. Supports dedup_by and min_module_size options for controlling how genes assigned to multiple modules are deduplicated and how small modules are filtered.
  • load_prebuilt_wss(), list_prebuilt_wss(), and wss_prebuilt_terms() provide access to a bundled reference module/weighting table ("core_AD_plasma_biomarkers") and its module term labels.

Prediction from core biomarkers

  • predict_WSS() predicts all 75 module scores for the "core_AD_plasma_biomarkers" reference set from Age, Gender, and 8 core plasma biomarkers, using a bundled joinet two-layer elastic net model, without requiring the full proteomic panel.
  • normalize_wss_biomarkers() builds the rank-normalized biomarker inputs predict_WSS() requires from raw biomarker values, via either method = "project" (projecting onto a bundled, anonymized baseline reference distribution, using project_rank_norm() – supports a single new patient) or method = "self" (rank-normalizing within your own cohort).
  • project_rank_norm() and load_prebuilt_biomarker_reference() expose the underlying projection utility and reference distribution directly.

The bundled model and reference data were verified end-to-end against the original analysis’s held-out follow-up cohort, reproducing published predictions to floating-point precision.