Phasing and Imputation

Haplotype estimation (phasing) and imputation.

Involves:

• Phasing: With reference and without reference
• Custom panel creation
• Imputation

Phasing tools implemented/to be implemented:

• EAGLE2
• SHAPEIT4
• BEAGLE5

Imputation tools implemented/to be implemented:

• Minimac4
• IMPUTE2
• BEAGLE5

Custom panel creation uses

• Minimac3

Uses bcftools throughtout and bgzip and plink2 occasionally.

Installing and using the workflow

Clone the workflow to your computer or cluster.

$ git clone https://github.com/esohkevin/siaImputationService.git

---

There are three main configuration scripts that need to be editted before running the workflow

1. containers.config

The script is located in the configs directory. Edit the following line by replacing "${HOME}/singularity" with your own path

cacheDir = "${HOME}/singularity"

This allows you to choose a location where the containers that are required will be stored. It is best to pick a location with enough storage capacity as the total size of all the containers could be upto 5GB

2. pbspro.config

The script is located in the configs directory. Edit this script to suit your cluster options.

executor {
    name = 'pbspro'
    queue = 'normal'                                            // replace 'normal' with your cluster queue option
    queueSize = 10                                              // increase or decrease the queue size according to your privileges 
}

...

process {
    beforeScript = 'module load chpc/singularity/3.5.3'         // load singularity according to your cluster directives
    cpus = 24
    time = 3.h
    clusterOptions = '-P XXXXXXX'                              // replace 'XXXXXXX' with the project name you are assigned to on your cluster
}

3. reference-hg*.config

There are two config files for reference sequences/panels: reference-hg19.config and reference-hg38.config

The files are located in the configs directory. Edit them accordingly to point the workflow to the appropriate reference sequence/panels location

For instance, for reference-hg19.config

params {
    panel_dir = ''			// directory where imputation reference panels are stored or will be stored
    ref_dir = ''			// directory containing fasta reference sequence
    fastaRef = ''			// name of the fasta reference sequence
    buildVersion = 'hg19'		// DO NOT EDIT THIS LINE!!!
}

The panel directory should contain subdirectories with different reference data types. See the following folder structure

referencePanels
├── hapmap
│   ├── genetic_map_chr1_combined_b37.txt
│   ├── ...
│   ├── genetic_map_chr22_combined_b37.txt
├── kgp
│   ├── chr1.1kg.phase3.v5a.vcf.gz
│   ├── chr1.1kg.phase3.v5a.vcf.gz.tbi
│   ├── ...
│   ├── chrX.1kg.phase3.v5a.vcf.gz
│   └── chrX.1kg.phase3.v5a.vcf.gz.tbi
├── plinkmap
│   ├── plink.chr1.GRCh37.map
│   ├── ...
│   └── plink.chrX.GRCh37.map
├── shapeit
│   ├── chr1.b37.gmap.gz
│   ├── chr1.b38.gmap.gz
│   ├── ...
│   ├── chrX.b37.gmap.gz
│   ├── chrX.b38.gmap.gz
└── tables
    ├── genetic_map_1cMperMb.txt
    ├── genetic_map_hg17_withX.txt.gz
    ├── genetic_map_hg18_withX.txt.gz
    ├── genetic_map_hg19_withX.txt.gz
    └── genetic_map_hg38_withX.txt.gz

The above three scripts are only edited once to set up the workflow on a new cluster.

The following script is the job configuration script which will be updated each time a different job is to be run

4. nextflow.config

Edit the following according to you input paramters.

params {
    phase = true
    with_ref = false
    impute = false
    phase_tool = 'shapeit4'			// options: shapeit4, eagle2, beagle5. [default: shapeit4] 
    impute_tool = 'minimac4'			// options: minimac4, impute2, beagle5. [default: minimac4]

    vcf = ''		    			// insert the VCF file to process here, with the full (absolute) path.
    out_dir = ''			    	// full path where the results will be saved
    out_prefix = ''				// base name of all output files

    RefereceSamplesToExclude = ''

    burninIter = 2000                   	// BEAGLE5 (this is only applicable to BEALGE5)
    mainIter = 1000                     	// BEAGLE5 (this is only applicable to BEALGE5)
    kpbwt = 20000                       	// BEAGLE5 | EAGLE2 (this is applicable to both EAGLE2 and BEAGLE5)
    pbwt = 8                            	// SHAPEIT4 (this is only applicable to SHAPEIT4)
    autosome = false

    // CREATING IMPUTATION PANEL
    input_dir = ''                      	// directory containing phased VCF files (single chromosomes) for which imputation panel is to be created
}

The following tutorial will describe how to use the nextflow.config file to run jobs

---

A. PHASING WITHOUT REFERENCE (NO IMPUTATION) The parameter scope should look like this

phase = true
with_ref = false
impute = false
phase_tool = 'shapeit4'
impute_tool = 'minimac4'

Then run the workflow as follows

nextflow run phaseAndImputeGenotypes.nf -w /path/to/work_directory/ -profile pbspro,hg19

• -w: this is the path where all temporary files will be stored. It should prefarably be a location with enough storage capacity
• -profile: the profile selector is added to make it easy to run the workflow on any cluster/system and using reference panels in different genome builds. Therefore, if your data set is in GRCh37 coordinate (hg19 or b37), then select hg19, etc.

---

B. IMPUTE GENOTYPES ONLY

This requires pre-phased data in single chromosome VCF files, bgzipped and in a single directory. The directory must also contain the tabix indexed for each VCF file.

The directory of the pre-phased data is supplied using input_dir in the nextflow.config file.

Example

chr1.vcf.gz
chr1.vcf.gz.tbi
.
.
.
chrX.vcf.gz
chrX.vcf.gz.tbi

NB: The directory must only contain the VCF files to be process (and their indexes) and must not contain any other files ending in '.vcf.gz'

---

C. PHASING WITHOUT REFERENCE AND THEN IMPUTE GENOTYPES

Make the following change in the nextflow.config script impute = true then run the workflow

nextflow run phaseAndImputeGenotypes.nf -w /path/to/work_directory/ -profile pbspro,hg19