LDLR Variant Prioritization Pipeline

High-Throughput Computational Reclassification of Variants of Uncertain Significance (VUS) in Familial Hypercholesterolemia

Preprint: Nasir, M. (2026). A High-Throughput Programmatic Pipeline for Computational Reclassification of Variants of Uncertain Significance in LDLR: Structural Evidence for Disruption of the LDL-Receptor Class A Ligand-Binding Repeats. Zenodo. https://doi.org/10.5281/zenodo.23323333

---

Overview

Familial Hypercholesterolemia (FH) is a common, severe autosomal dominant genetic disorder caused primarily by mutations in the low-density lipoprotein receptor (LDLR) gene. Clinical sequencing frequently uncovers Variants of Uncertain Significance (VUS), leaving patients without actionable diagnostic insights.

This repository houses a fully programmatic, reproducible command-line interface (CLI) pipeline designed to bypass manual web-based databases. By integrating genomic extraction, programmatic batch API annotation, multi-parameter clinical filtering, and 3D structural biophysics, this pipeline systematically reclassifies LDLR VUS into high-confidence pathogenic candidates.

---

Pipeline Architecture

The workflow is managed via a fault-tolerant Bash orchestrator (run_pipeline.sh) executing five core programmatic modules:

1. Genomic Data Extraction — Isolates single-nucleotide VUS from raw ClinVar VCF files using bcftools.
2. High-Throughput Batch Annotation — Interfaces with the Ensembl REST Protein–Ensembl API (/vep/human/region) via Python (requests) with robust retry logic to extract evolutionary (SIFT), biophysical (PolyPhen-2), and population frequency (gnomAD) metrics.
3. Strict Pathogenicity Filtering — Applies mathematical cutoffs (gnomAD < 0.0001, SIFT < 0.05, PolyPhen-2 > 0.90) via pandas to isolate top-tier candidates.
4. Functional Domain Mapping — Extracts canonical transcript nomenclature (HGVS.p) and structural database classifications (InterPro/Pfam).
5. Structural Validation — Dynamically queries the AlphaFold API to retrieve the latest structural model (.cif) for PyMOL visualization.

---

Repository Structure

├── run_pipeline.sh          # Master Bash orchestrator script
├── annotate_gene.py         # Batch Ensembl REST API annotation module
├── rank_variants.py         # Multi-parameter clinical filtering module
├── map_domains.py           # Canonical transcript & HGVS domain mapping module
├── download_structure.py    # Dynamic AlphaFold CIF fetcher module
├── top_LDLR_candidates.csv  # Mathematically prioritized variant dataset
├── figure1_ldlr_final.png   # 3D PyMOL structural visualization figure
└── README.md                # Project documentation

---

Prerequisites & Dependencies

Ensure your environment (WSL Ubuntu or Linux) satisfies the following requirements:

• Core Tools: bash, git, bcftools
• Python Environment (v3.10+): pandas, requests, argparse
• Visualization: PyMOL (v3.x)

---

Step-by-Step Execution Guide

1. Clone the Repository

git clone https://github.com/mubeeen1/LDLR-Variant-Prioritization.git
cd LDLR-Variant-Prioritization

2. Prepare the Input Data

Download the latest human ClinVar VCF and its index file into your root working directory:

# Example paths for ClinVar GRCh38 VCF
wget https://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/clinvar.vcf.gz
wget https://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/clinvar.vcf.gz.tbi

3. Run the Master Orchestrator

Execute the complete end-to-end workflow with a single command:

chmod +x run_pipeline.sh
./run_pipeline.sh clinvar.vcf.gz LDLR 19 P01130

---

Results & Biological Interpretation

Running the pipeline on the LDLR gene (Chr 19, UniProt: P01130) yielded the following structural and evolutionary insights:

• Data Funnel — Out of 1,187 ClinVar LDLR variants of uncertain significance, 1,101 valid single-nucleotide variants were annotated via Ensembl VEP; joint filtering (gnomAD < 0.0001, SIFT < 0.05, PolyPhen-2 > 0.90) isolated 222 high-confidence pathogenic candidates.
• Structural Impact — Top variants (p.Cys132Trp, p.Asp143His, p.Cys149Tyr, p.Cys161Gly, p.Cys214Ser) cluster inside the extracellular LDL-receptor class A repeats (3–5) of the ligand-binding domain.
• Disulfide Network Disruption — Four of the five top candidates substitute conserved cysteine residues (Cys132, Cys149, Cys161, Cys214), destroying the disulfide-bond scaffolding that stabilizes the ligand-binding repeats and explaining the structural mechanism of pathogenicity.

3D Structural Validation (PyMOL)

To render the publication-grade structural model displaying the disrupted cysteine–disulfide bonding network:

load LDLR_structure.cif, ldlr
bg_color white
hide everything
show cartoon
color gray70, ldlr
select pathogenic_vus, resi 132+143+149+161+214
show sticks, pathogenic_vus
color red, pathogenic_vus
select disulfides, resname CYS and resi 130-220
show sticks, disulfides
color yellow, disulfides
color red, pathogenic_vus
zoom pathogenic_vus, 15
ray 1600, 1200
png figure1_ldlr_final.png, dpi=300

---

License & Author

Developed as an open-source, reproducible bioinformatics pipeline suitable for academic research and clinical genomics variant interpretation portfolios.

• Author: [Mobeen Nasir / mubeennasir117@gmail.com]
• Target Condition: Familial Hypercholesterolemia (CDC Tier-1 Genomic Application)