PIE

PIE (Phasing all-In-one Evaluator) is a command-line tool for evaluating haplotype phasing.

Installation

Currently, the prebuilt wheel supports Python 3.8 on Linux x86_64.

Option1. Install through pip

Due to name conflict on PyPI, we have to use pie-phasing-eval for pip installation. It is just a name change, won't affect the command line.

conda create -n pie python=3.8;
conda activate pie;
pip install pie-phasing-eval;

Option2. Install through Conda

conda create -n pie python=3.8;
conda activate pie;
conda install -c bioconda pie;

Option3. Local installation

conda create -n pie python=3.8;
conda activate pie

# Download and extract PIE, then enter the directory containing setup.py
pip install .;

Usage

Required arguments

Parameters	Description	Format or example
-i or --input	file for evaluation	.vcf / .vcf.gz / .bcf
-c or --compare	ground truth file	.vcf / .vcf.gz / .bcf
-r or --ref	reference file	.fa / .fasta / .fai
-o or --output	Output file prefix	exmaple: ./test/output_name

Recommend flag

Use --verbose to monitor the running process and enable detailed logging.

Full usages

Usage: pie [OPTIONS]

Options:
  -i, --input TEXT       Input vcf/vcf.gz file for evaluation  [required]
  -n, --name TEXT        User defined sample name, [default: Sample]
  -c, --compare TEXT     Ground truth vcf/vcf.gz file for comparison
                         [required]
  -r, --ref TEXT         Reference file fasta file (.fasta or .fa) or fasta
                         index file (.fai)  [required]
  -o, --output TEXT      Output file prefix,  such as -o ./test/output_name
                         [required]
  -t, --threads INTEGER  Maximum number of parallel threads [default: 24]
  -m, --max-len INTEGER  Maximum variant distance for pairwise calculation
                         [default: 2473538]
  -b, --bed TEXT         .bed file specifying genomic regions to include
                         [default: None]
  --min-sv INTEGER       Minimal length threshold of Structural Variant
                         [default: 30, ALT length > 30 bp is SV]
  --chrom TEXT           Chromosome to evaluate,use comma to join chromosome
                         name e.g. --chrom chr1,chr2,chr3
                         [default:chr1,chr2,chr3,...,chr22]
  --sexchrom TEXT        Sex chromosome,use comma to join chromosome name e.g.
                         --sexchrom chrX,chrY [default: chrX,chrY]
  --mincount INTEGER     Minimum number of phased sites in a phase block
                         [default: 2]
  --block                Output phasing block start and end positions in a BED
                         file
  --no-sex               Ignore sex chromosome
  --canonical            Canonical mode, only evaluate single mutation SNV
                         ignore double heterozygous site
  --only-snv             Only evaluate single nucleotide variation
  --only-indel           Only evaluate insertion and deletion
  --only-sv              Only evaluate structural variant
  --no-snv               Ignore single nucleotide variation
  --no-indel             Ignore insertion and deletion
  --no-sv                Ignore structural variant
  --no-double            Ignore double heterozygous site
  --no-sort              Do not sort chromosome or regions, directly use the
                         input order
  --verbose              Enable verbose mode, printing parameters and progress
                         to standard output
  --version              Show the version and exit.
  --help                 Show this message and exit.

Examples

Suppose phase.vcf is the sample VCF file to be evaluated against the ground truth VCF file (truth.vcf).

1. (Comprehensive) Evaluate all autosomes

pie --verbose -i phase.vcf -c truth.vcf -r ref.fa -o ./output/comprehensive

2. (Chromosome specific) Restrict evaluation to selected chromosome

pie --verbose --chrom chr6 -i phase.vcf -c truth.vcf -r ref.fa -o ./output/chr6

3. (Region specific) Focus only on regions defined in a BED file.

pie --verbose -i phase.vcf -c truth.vcf -r ref.fa --bed mhc.bed -o ./output/mhc

4. (Filter) Exclude Structural Variants (SV) from the analysis

pie --verbose -i phase.vcf -c truth.vcf -r ref.fa --no-sv -o ./output/no_sv

5. (More output) Output raw phasing block start and end positions to a BED file

pie --verbose -i phase.vcf -c truth.vcf -r ref.fa --block -o ./output/more

6. (Miscellaneous) Evaluate only SNVs on chromosome 6 and output raw block start and end positions to a BED file.

pie --verbose -i phase.vcf -c truth.vcf -r ref.fa --chrom chr6 --only-snv --block -o ./output/misc

Test data

Small query and truth test files are provided in here including three formats:

Filename	Format	Description
small_query.vcf	VCF	Uncompressed VCF
small_query.vcf.gz	VCF.GZ	Compressed VCF
small_query.bcf	BCF	Binary VCF

small_truth.vcf is provided as the truth.

The VCF format follows the VCF v4.1 specification https://samtools.github.io/hts-specs/VCFv4.1.pdf

Output file

Output suffix	Description	Condition
.variant.stats.csv	Genotype evaluation result and phased percentage	Always generated
.perchrom.csv	Per chromosome phasing evaluation result	Always generated
.overall.csv	Overall sample evaluation result	Always generated
.blocks.bed	Raw phasing block start and end in -i or --input file	with --block set

Resource consumption

Pie is expected to complete evaluation within minutes with the default 24 threads on an x86_64 platform.

The actual performance may vary depending on factors such as size of vcf, I/O speed, memory speed, and CPU capabilities.

Acknowledgements

Pie is dependent on the following libraries, we are grateful to all the developers/maintainers:

• click: Python command line
• cyvcf2: VCF/BCF processing
• pyfastx: Reference FASTA processing
• numba: JIT acceleration
• sortedcontainers: Python Sorted Container Types