hcCNA-bench and hcCNA-bench-tree: Benchmark Datasets for Evaluating Haplotype- and Clone-Specific Detection of Complex Copy Number Alterations in Single-Cell DNA

This repository contains hcCNA-bench and hcCNA-bench-tree, the complete benchmark datasets used in our study: "Benchmarking haplotype- and clone-specific detection of complex copy number alterations in single-cell DNA with HCDSIM and HCBench" We deposit the full benchmark here so that the results reported in the paper are fully reproducible and reusable. The archive includes all datasets used in the manuscript, specifically: All ground-truth profiles — the per-cell, haplotype- and clone-specific copy number annotations generated with the exact configuration and random seeds used in the paper. All simulated sequencing reads — provided as barcoded, single-cell resolved BAM (stored as CRAM, see below) files for every simulated dataset. Together these allow other researchers to run their own methods against exactly the same data used in our benchmark. The simulator used to generate these datasets, HCDSIM, is available on GitHub at https://github.com/BioThinkLab/HCDSIM, with full documentation at https://hcdsim.readthedocs.io. 1. Merging the split archive Because the full dataset (~145 GB) exceeds the single-file upload limit, the original archive HCDSIM-Datasets.zip was split into 30 parts (HCDSIM-Datasets.zip.part-00 … HCDSIM-Datasets.zip.part-29) of ~5 GB each. After downloading all parts into the same directory, reconstruct and extract the original archive with: # 1. Concatenate all parts back into the original zip cat HCDSIM-Datasets.zip.part-* > HCDSIM-Datasets.zip # 2. Unzip unzip HCDSIM-Datasets.zip You can verify the integrity of each part against the MD5 checksums listed for every file on this Zenodo record before merging. 2. Converting CRAM back to BAM To save storage space, all simulated single-cell read files were converted from BAM to CRAM using samtools. CRAM is a reference-based compressed format, so you will need the same reference genome that was used during simulation to decode the reads. All datasets were simulated against the human reference genome hg38 (hg38.fa), which can be downloaded from the UCSC Genome Browser. Make sure the FASTA is indexed (samtools faidx hg38.fa) before use. To convert a single CRAM file back to BAM: # Convert a single CRAM to BAM samtools view -b -T hg38.fa -o sample.bam sample.cram # Index the resulting BAM samtools index sample.bam To convert many CRAM files in a directory at once: for f in *.cram; do samtools view -b -T hg38.fa -o "${f%.cram}.bam" "$f" samtools index "${f%.cram}.bam" done 3. Ground-truth profiles Detailed descriptions of the ground-truth profile file formats and columns (haplotype-specific and clone-specific copy number annotations) are provided in the HCDSIM documentation: ➡️ https://hcdsim.readthedocs.io Please refer to the documentation for the exact schema of each profile file and guidance on interpreting the haplotype- and clone-level annotations.