Skip to content

Repository files navigation

CERFAC

Cloud-based, Evidence-based Rigorous Functional Assay Calibration

A bioinformatics pipeline for validating functional assay predictions against clinical observational data using WDL workflows on Terra or locally.

Quick Start

  1. New to CERFAC? Read CERFAC_documentation.md for the full step-by-step guide
  2. Setting up locally? See CROMWELL_SETUP.md for workflow execution
  3. Want to understand the architecture? Check CLAUDE.md for detailed technical information

What It Does

CERFAC combines three data sources to calibrate functional assay predictions:

  • ClinVar: Clinical variant classification from NCBI
  • gnomAD: Population variant frequency data
  • Your assay scores: Functional predictions you provide

The pipeline normalizes variants across formats, merges the data, and performs statistical calibration (OR calculations) via Jupyter notebook.

Running on Terra (Production)

See CERFAC_documentation.md — complete step-by-step guide for cloning a workspace and running workflows.

Running Locally (Development)

See CROMWELL_SETUP.md for workflow execution or CLAUDE.md for technical details.

Architecture

ClinVar Query → ┐
                ├→ Merge & Normalize → Statistical Analysis
gnomAD Query  → ┤   (VRS API)          (R Jupyter notebook)
Your Data ────→ ┘

Three Docker images handle different stages:

  • cerfac-clinvar — NCBI EDirect extraction
  • cerfac-merge — Data merging and API calls
  • cerfac-gnomad — Population frequency lookup via Hail

Documentation

Document Purpose
CERFAC_documentation.md User guide for Terra
CLAUDE.md Developer reference
CROMWELL_SETUP.md Local workflow execution
docker/README.md Docker image building

Key Features

  • Variant normalization: Handles HGVS coding, HGVS genomic, and gnomAD VCF formats
  • Scalable: Runs on Terra cloud or local machine via Cromwell
  • Reproducible: WDL workflows with containerized dependencies
  • Flexible: Accepts TSV, CSV, or TXT input files

Requirements

  • Docker (for containerized task execution)
  • Java Runtime (for Cromwell workflow engine)
  • Sufficient disk space (~2 GB for Docker images)

Questions?

Refer to the documentation files above or check CLAUDE.md for technical details.

About

Cloud-based, Evidence-based Functional Assay Calibration

Resources

Stars

1 star

Watchers

5 watching

Forks

Releases

Packages

Contributors

Languages