somalier

somalier estimates relatedness and verifies sample identity by extracting compact genome sketches from BAM/CRAM/VCF/GVCF files to detect sample swaps and relatedness in cancer and germline sequencing studies.

Key Features:

• Efficient genome sketching: Operates directly on alignment files (BAM/CRAM) and VCF/GVCF without requiring jointly-called germline variants, extracting compact sketches of informative genetic variation.
• Speed and scalability: The somalier relate command executes in under two seconds for 600 samples and approximately one minute for 4,500 samples, enabling rapid comparison across hundreds to thousands of samples.
• Compatibility with data types and genome builds: Supports BAM, CRAM, VCF, GVCF and multiple genome builds for diverse sequencing datasets.
• Tumor-normal and somatic contexts: Designed to detect sample swaps and verify relatedness in tumor versus matched normal comparisons and studies reporting primarily somatic variants.
• Pedigree and cohort comparison: Compares sample sketches across cohorts and to pedigree files to identify related samples and sample identity issues.

Scientific Applications:

• Cancer genomics: Ensures sample integrity in tumor/normal, spatial, longitudinal, and cell-free DNA sequencing studies where germline joint calls may be unavailable.
• Germline studies: Provides rapid relatedness checks and pedigree verification across large cohorts.
• Demonstrated datasets: Applied to five glioma sample sets (normal, tumor, and cell-free DNA) and high-coverage 1000 Genomes Project data to identify related samples.

Methodology:

Extracts compact genome sketches by identifying informative sites from BAM/CRAM/VCF/GVCF and compares sketches using the somalier relate command to compute relatedness without requiring joint germline variant calling.