kilobase-resolution-genomics-analysis

kilobase-resolution-genomics-analysis

Summary

Apply Juicer's integrated pipeline and command-line tools to process raw Hi-C sequencing data (FASTQ) into kilobase-resolution contact maps (.hic files) and annotate structural features such as loops and topologically associating domains (TADs). This skill encompasses both the primary data processing stage (alignment, deduplication, .hic file generation) and post-processing feature annotation.

When to use

You have raw Hi-C FASTQ data and need to generate contact maps at kilobase resolution, or you have pre-generated .hic files and need to annotate structural features (loops, domains) for downstream 3D genome analysis. This skill is appropriate when working with chromosome conformation capture (3C-based) experiments and requires parallel cluster or cloud infrastructure.

When NOT to use

• Input data are not Hi-C or 3C-based; Juicer is specialized for chromosome conformation capture.
• You already have pre-processed, normalized contact matrices in a standard format (e.g., HDF5, sparse matrix) and only need downstream statistical analysis, not raw-read processing.
• Your cluster or compute environment does not support any of the supported job schedulers (SLURM, LSF, GridEngine, OpenLava) or cloud/local execution; Juicer requires orchestration infrastructure.

Inputs

• FASTQ files (raw Hi-C sequencing reads)
• Genome reference (FASTA or genome ID, e.g., 'hg19')
• Restriction enzyme site file (coordinates of cut sites)
• Chromosome sizes file (.chrom.sizes)

Outputs

• .hic contact map file (multi-resolution, indexed)
• Annotated loop coordinates (BED/BEDPE format or Juicer native)
• Topologically associating domain boundaries (domain annotation file)
• Pipeline statistics and QC logs

How to apply

The Juicer platform operates in two sequential stages. First, invoke the main pipeline (juicer.sh) on your cluster with FASTQ files located in [topDir]/fastq, specifying the genome ID (e.g., 'hg19' or 'mm10'), restriction enzyme site (e.g., 'HindIII', 'MboI'), and resource queue parameters appropriate to your scheduler (SLURM, LSF, GridEngine, or single CPU). The pipeline performs alignment via BWA, chimera handling, deduplication, and generates a merged .hic contact map file. Second, execute post-processing feature annotation tools (accessible via Juicer command-line tools) on the resulting .hic file to call loops (using HiCCUPS, which requires CUDA/GPU for optimal performance) or detect topologically associating domains. Validate output by checking that the .hic file is properly indexed, contains multi-resolution matrix data at 5 kb and coarser scales, and that annotation outputs (loop coordinates, domain boundaries) are in standard formats (e.g., BED, BEDPE). The choice between GPU-accelerated and CPU-only HiCCUPS depends on available hardware; CPU versions are available but slower.