Zebrafish Variant Calling Training Project

Image by Petr Kuznetsov from Pixabay

🐟 Variant Calling in Zebrafish

In this post, I document a hands-on training project I conducted using real sequencing data from the zebrafish study PRJNA630223. This study from University of Glasgow investigated how fisheries-induced selection —specifically simulated trawling— affects the genetic landscape of zebrafish, with a focus on how population density modulates these responses.

📚 The original study involved over 90 zebrafish samples, reared under different density conditions, gender and subjected to repeated harvest. However, my project only analyzes four samples for training purposes.

🧪 Objective

The goal of this project was to practice and understand the steps involved in a variant calling pipeline, from raw sequencing reads to functional variant annotation.

🧬 Data Used

I selected four male zebrafish samples, all from the baseline density group:

• 2 “Captured” fish: caught during simulated trawling
• 2 “Escaped” fish: those that evaded capture

This distinction is relevant because the study explores how behavioral traits like sociality or boldness may lead to capture-associated selection, potentially leaving genomic signatures that differ between groups.

📝 Sample details and Accession IDs:

• Captured1: SRR11676923
• Captured2: SRR11676919
• Escaped1: SRR11676853
• Escaped2: SRR11676862

🔬 Pipeline Overview

All steps were conducted using open-source tools within a Galaxy + Jupyter environment:

1. Data download from ENA
2. Quality control using FastQC & MultiQC
3. Trimming low-quality reads using fastp
4. Alignment with minimap2
5. SAM/BAM processing using samtools
6. Variant calling with freebayes
7. Filtering based on quality (QUAL ≥ 20) using vcffilter
8. Annotation using Ensembl’s VEP with a custom zebrafish GTF using Galaxy EU

📊 Results

Note: These results are purely descriptive and reflect only the four samples I selected for training. They’re not representative of the full study and aren’t meant to support any biological claims — just to help me learn from real data.

1. Total Variants Per Sample

This bar plot shows the number of filtered variants (after QUAL ≥ 20) in each sample.

As seen above, escaped fish tended to have more total variants than captured ones.

2. Variant Impact Summary

Each variant was classified into one of four impact levels by VEP: HIGH, MODERATE, LOW, or MODIFIER.

The plot below compares their distribution across samples.

The plot excludes MODIFIER consequences, which dominate most genomes but are less informative in this training context and may obscure other types.

While LOW-impact variants dominate across all samples, subtle differences exist in the counts of HIGH and MODERATE variants.

3. Variant Summary Table

📌 All counts based on filtered variants only.

🧠 What I Learned

This project helped me solidify several key concepts in variant analysis:

• 🛠️ Practical tool use: I became more comfortable using minimap2, freebayes, samtools, and VEP in a real workflow.
• 📈 Data interpretation: I learned how to read and summarize variant call statistics meaningfully.
• 🎨 Visualization skills: I explored how to represent genomic differences clearly using plots.
• 🧭 Confidence for future work: This training project gave me a clear roadmap for scaling up future analyses, whether for zebrafish, humans, or microbes.

📂 Code & Pipeline

You can find the pipeline notebook and documentation here:
🔗 GitHub Repo Link

📚 Main Study

The Original study’s datasets can be accessed here:
🔗 ENA Project PRJNA630223

Original study Citation:

• Crespel A, Schneider K, Miller T, et al. Genomic basis of fishing-associated selection varies with population density. Proc Natl Acad Sci U S A. 2021;118(51):e2020833118. doi:10.1073/pnas.2020833118