SAFARI-IGHJ: From Computational Prediction to In Vivo Proof

Large-scale validation of IGHJ gene functionality across 13 wild ruminant species and 7 Bovidae tribes

Overview

This repository contains the source code, validation scripts, and analysis pipelines for the SAFARI-IGHJ manuscript. SAFARI-IGHJ is a species-agnostic bioinformatics pipeline for discovering immunoglobulin heavy chain J gene segments (IGHJ) from whole-genome assemblies of non-model ruminants.

Repository Structure

SAFARI_Paper_Release/
├── SAFARI_Manuscript_FinalDraft.md   # Full manuscript text
├── references.bib                     # BibTeX references
├── figures/
│   ├── fig1_validation_levels.pdf     # Evidence levels by tribe
│   ├── fig2_splicing_reads.pdf        # RNA-seq mapped vs spliced reads
│   └── graphical_abstract_prompt.txt  # Prompt for graphical abstract generation
├── scripts/
│   ├── fig1_validation_levels.py      # Figure 1 generation script
│   └── fig2_splicing_reads.py         # Figure 2 generation script
├── supplementary/
│   └── (supplementary tables and notes)
├── .zenodo.json                       # Zenodo metadata for DOI minting
└── README.md                          # This file

Pipeline Components

SAFARI-IGHJ v1 & v2 (Discovery)

• tBLASTn mining with relaxed parameters for short, divergent IGHJ segments
• Single-linkage locus clustering to separate genuine IGHJ clusters from pseudogene noise
• RSS-23 information content (IC) scoring
• FR4 motif classification

SAFARI-IGHJ-Heavy (Validation Suite)

• 11 validation subcommands, 33 unit tests
• Positive/negative controls, parameter robustness (36 combinations)
• Fragmentation stress testing (recall = 1.0 at N50 >= 100 kb)

SAFARI-Score (ML Classifier)

• sklearn-based classifier for candidate ranking
• Leave-one-species-out cross-validation (LOOCV)
• Composite score: 0.881 (gold F1 = 1.0, LOOCV mean = 0.898)

V(D)J Validation Framework (3 Pillars)

• Pillar 1: Transcriptional activity (RNA-seq mapping)
• Pillar 2: JH-to-CH1 splicing (CIGAR N operations)
• Pillar 3: Somatic V(D)J recombination (WGS soft-clip analysis)

Key Results

• 21 species analyzed across 9 taxonomic groups
• 13 species validated with multi-omics data
• >11.4 million RNA-seq reads mapped to predicted IGHJ loci
• >1.49 million splice-junction reads confirming mRNA maturation
• >1,400 V(D)J junction soft-clips from 30+ wild individuals
• 7/7 Bovidae tribes validated (100% tribal coverage)

Requirements

• Python >= 3.10
• BLAST+ >= 2.13
• minimap2 >= 2.26
• samtools >= 1.17
• scikit-learn >= 1.3
• matplotlib, seaborn, pandas, BioPython

Citation

If you use SAFARI-IGHJ in your research, please cite:

Correia, J.P. (2026). From Computational Prediction to In Vivo Proof: Large-Scale Validation of SAFARI-IGHJ Across 11.4 Million Reads Confirms Immunoglobulin J Gene Functionality in Wild Ruminants. BMC Genomics [submitted].

License

MIT License

Contact

Jean Pierre Correia — jpierre.vd@gmail.com | ORCID: 0009-0004-3566-3987