Computational Reclassification of MLH1 Variants of Uncertain Significance (VUS)

Project Overview

This repository contains an automated bioinformatics pipeline designed to extract, filter, and computationally annotate missense Variants of Uncertain Significance (VUS) within the MLH1 gene.

Germline loss-of-function mutations in MLH1 drive DNA mismatch repair (MMR) deficiency, serving as the primary genetic susceptibility mechanism for Lynch syndrome and hereditary colorectal cancer. This project specifically targets the highly conserved N-terminal ATPase domain (amino acids 1–330), which is critical for ATP binding and hydrolysis during the mismatch repair process.

Computational Methodology

The pipeline is engineered in Python and executes a stepwise integration of clinical and genomic databases via REST APIs.

1. Clinical Extraction (01_variant_extraction.py)

• Source: NIH ClinVar API (via Bio.Entrez).
• Function: Dynamically queries ClinVar for all recorded missense variants within MLH1 currently classified strictly as "Uncertain Significance."
• Output: Generates a raw summary dataset of unclassified clinical mutations.

2. Domain Targeting (02_domain_filtering.py)

• Function: Utilizes Regular Expressions (Regex) to parse HGVS protein nomenclature and isolate mutations located exclusively within the target catalytic boundaries (AA 1–330).
• Output: Formats a refined subset of variants for downstream algorithmic prediction.

3. Algorithmic Annotation (03_api_annotation.py)

• Source: MyVariant.info API.
• Function: Programmatically queries aggregated variant databases to append predictive meta-scores and population metrics to each ClinVar ID.
• Key Metrics Retrieved:
  • CADD (Combined Annotation Dependent Depletion): Assesses the deleteriousness of single nucleotide variants (scores > 20 indicate the top 1% most deleterious mutations).
  • REVEL (Rare Exome Variant Ensemble Learner): An ensemble method predicting the pathogenicity of missense variants.
  • gnomAD Exome Allele Frequency: Validates variant rarity against global healthy population sequencing data.

Current Project Status

• Pipeline Status: Complete and automated.
• Data Yield: Successfully extracted and annotated 445 missense VUS within the MLH1 ATPase domain.
• Next Steps: Statistical triage of the annotated evidence table to identify highly discordant predictions and isolate severe "Hidden Pathogenic" candidates for cross-referencing with published functional in-vitro assays.

Author

Barada Chakraborty Interdisciplinary Researcher | Systems Immunology & Computational Genetics