In clinical or population genomics, it is often neccessary to check sample identity in a fast, reliable and privacy-preserving way. The purpose of this project is to generate a compact, privacy-aware representation of an individual's genotype using a selected set of informative SNPs.
Our method fully obscures the underlying genotypes, making it completely impossible to infer them even in trios.
The resulting hash string of a sample can be output as a 2D barcode to be included in lab reports.
For working with human data, we also generated an optimal set of 462 biallelic SNPs that are highly informative in all major populations and located in exonic regions. This universal SNP set can be used as an efficient genotyping panel for both WES and WGS.
The variant genotypes are encoded as semi-randomly picked pairs in such a way that the resulting value, while retaining some information about the variants' relative states within the pair, always allows for any single genotype (0/0, 0/1 or 1/1) to occur at each of the positions.
Before using the tool, you must generate a VCF file from your BAM (or CRAM) file using a predefined set of filtered variant positions. Use the following command:
bcftools mpileup -Ou \
-R <snp_regions.bed> \
-f <reference.fasta> <input.bam> | \
bcftools call -A -C alleles \
-T <snps.tsv.gz> \
-i -m -Ov > <output>.vcf-
Input CRAM/BAM files must be aligned to the GRCh38 (hg38) reference genome.
-
The following files with variant positions/alleles and the surrounding regions are provided in the cloud link below: gnomad.exomes.v4.1.sites.distancefilt500k.ALL.tsv.gz gnomad.exomes.v4.1.sites.distancefilt500k.ALL.regions.bed
-
Replace the placeholders with appropriate files:
<reference.fasta>— the GRCh38 reference genome FASTA file.<regions.bed>— a BED file with intervals surrounding each SNP (provided)<input.bam>— your input CRAM or BAM file.<snps.tsv.gz>— a TSV file with SNP positions and alleles (provided)<output>.vcf— the name of the resulting VCF file.
The hash string for publicly available sample HG001 (NA12878) from 1000GP looks like this: c0b9b6a71999420ebbebe9669e6b7e266ab661a172795e9716e14b97bb119a1491976962990e6
To install the tool with required dependencies, run the following command from the root directory of the project:
pip install -e .The tool has two main functionalities, which can be invoked from the command line:
- Barcode Generation: To compute a barcode from a single VCF file.
- Barcode Comparison: To compare signatures between a VCF files or a previously generated string.
To generate a barcode from a VCF file, use the following command:
gbtools barcode <path_to_vcf_file> --output_filename <output_filename> --output_path <output_directory><path_to_vcf_file>— the path to the VCF file.--output_filename— (optional) the name of the output barcode image file.--output_path— (optional) the directory where the barcode image will be saved.
To compare the signature of a VCF file with a previously generated barcode, run:
gbtools compare <input1> <input2><input1>- VCF path or precomputed hash string<input2>- VCF path or precomputed hash string
The command will output the Hamming distance between the two hashes, as well as the fraction of non-missing hash positions being compared.
For testing the tool, you can use genomic and exomic data available data.
├── LICENSE
├── README.md
├── figs
│ └── SNP_encoding.png
├── genetic_barcode
│ ├── __init__.py
│ ├── cli
│ │ ├── __init__.py
│ │ └── main.py
│ ├── core
│ │ ├── __init__.py
│ │ ├── barcoding.py
│ │ └── comparing_hashes.py
│ └── utils
│ ├── __init__.py
│ └── hashing.py
├── requirements.txt
└── setup.py
-
Giulgaz Muradova
GitHub | email: gm@clinbio.ru -
Fedor Konovalov
GitHub | email: fk@clinbio.ru