Benchmarking SNV callers with a consensus-based pseudo-truth set framework in the non-model organism Chelonia mydas.

This dataset supports the study titled “Benchmarking SNV callers with a consensus-based pseudo-truth set framework in the non-model organism Chelonia mydas.” The dataset contains whole-genome sequencing (WGS)-derived variant call sets used to evaluate single-nucleotide variant (SNV) detection in Chelonia mydas. The analysis was performed on high-depth (~60×) Illumina sequencing data from two individuals: a phenotypically normal sample (CH-NORMS1) and a fibropapillomatosis-affected sample (Ab-NormS2). Reads were preprocessed using fastp and aligned to the rCheMyd1.pri.v2 reference genome using BWA-MEM. SNVs were independently identified using five widely used variant callers: DeepVariant, GATK HaplotypeCaller, FreeBayes, VarScan, and BCFtools. Consensus-based pseudo-truth sets were generated by intersecting variant call sets across multiple thresholds (3, 3+, 4, 4+, and 5 callers), providing high-confidence variant subsets for benchmarking and downstream analysis. Performance evaluation (precision, recall, and F1-score) was conducted using RTG vcfeval, and concordance analysis was performed using Jaccard similarity metrics. 📂 Dataset Contents This dataset includes: Individual SNV call sets (VCF.gz) for each variant caller Corresponding index files (.tbi and .csi) Filtered SNV datasets (FILTER = PASS, AD > 3, VAF > 0.02) Consensus-based pseudo-truth sets for both CH-NORMS1 and Ab-NormS2 datasets Threshold-based truth sets (n = 3, 3+, 4, 4+, 5 callers) Compressed archives of truth sets for reproducibility 🔗 Raw Sequencing Data Availability The raw whole-genome sequencing (FASTQ) data used in this study are publicly available in the NCBI Sequence Read Archive (SRA) under: BioProject: PRJNA1312993 Samples: CH-NORMS1 (Normal) BioSample: SAMN50881756 SRA: SRS26352173 Ab-NormS2 (Rehabilitation pool individual) BioSample: SAMN50881755 SRA: SRS26352174 🔗 Code and Workflow Availability All benchmarking outputs, concordance analyses (including Jaccard similarity), and workflow scripts are publicly available at: 👉 https://github.com/FahadAslam988/chelonia-snv-benchmarking This repository contains: RTG vcfeval benchmarking results (precision, recall, F1-score) Jaccard similarity and concordance analysis outputs Variant filtering and processing scripts Complete reproducible SNV benchmarking workflow