PERADIGM

Phenotype Embedding Similarity-based Rare Disease Gene Mapping

This repository contains the R code supporting the analysis described in the paper:
PERADIGM: Phenotype Embedding Similarity-based Rare Disease Gene Mapping

PERADIGM is a framework that integrates phenotype embedding and patient similarity to identify rare disease-associated genes using large-scale biobank data. This repository includes code to replicate the key analyses and figures from the study.

• main.R: Main script for the analysis, including:

  • Data loading and preprocessing
  • Running phenotype-gene association tests
  • Generating similarity matrices and embeddings
  • Outputting statistical results

• function.R: Contains all helper functions for:

  • Embedding computation
  • Similarity scoring
  • Regression-based testing
  • Carrier/control selection

Place your data files using the following directory structure:

data/
├── R_doc/
│   ├── hesin_diag_all_new.RData
│   ├── eid_all.RData
│   ├── cov_adjust.RData
│   └── IC_hesin_500k.csv
├── icd_related/
│   └── ICD10_mapping.csv
├── generate_all_gene_pos/
│   └── gene_info.RData
├── embedding/
│   └── hesin_icd10_descrip_embed.txt
└── hesin_diag.txt   # Optional/redundant diagnosis file

To reproduce the analysis:

1. Ensure R and required packages are installed.
2. Place the data files in the correct subfolders as shown above.
3. Run main.R to initiate the pipeline.