Skip to main content Link Search Menu Expand Document (external link)

UK Biobank WGS data

Table of contents

  1. Rationale
  2. Phasing the WGS data
    1. Quality control
    2. Phasing
      1. Phasing common variants in chunks
      2. Ligate chunks
      3. Phasing rare variants in chunks
      4. Concatenate chunks
  3. Validation of your phasing

For any question on this pipeline, please contact Robin J. Hofmeister and Diogo M. Ribeiro.


Rationale

SHAPEIT5 is a two-step approach that treats each chromosome independently and works as follows:

  1. Phase common variants (MAF >= 0.1%) of a chromosome using SHAPEIT5_phase_common. This can be done as a single job for SNP array or WES data, but it might be necessary to split the chromosome into large chunks (e.g. 20 cM) in the case of WGS data. Chunk coordinates for b38 are given in the resources folder.

  2. Ligate the phased common variants (MAF >= 0.1%) of a chromosome using SHAPEIT5_ligate, only if chunking was performed in step 1. The ligation step is computationally light and uses variants in the intersection of the chunks to provide chromosome-wide haplotypes. The result of this step (or the previous step if no chunking was used), is used as a haplotype scaffold for the next step.

  3. Phase rare variants (MAF < 0.1%) of a chromosome using SHAPEIT5_rare. To do this, we use the haplotype scaffold generated in step 2 (or 1) and we proceed in relatively small chunks (e.g. 5Mb) to run relatively fast job in parallel. At the end of this step, we have several fully phased chunks across the chromosome.

  4. Concatenate the phased chunks generated in step 3 using bcftools concat -n. As in the previous step haplotypes have been phased onto a haplotype scaffold, there is no need to ligate the chunks, and the files can be concatenated without decompression and recompression. This makes this step almost instantaneous, even for large cohorts.


Phasing the WGS data


If you want to phase the UK Biobank WGS data on the Research Analysis Platform (RAP) DNAnexus, you can directly use the pipeline provided in the github.

Clone the github repository using git clone https://github.com/odelaneau/shapeit5.git and navigate to the dedicated folder with cd shapeit5/tasks/phasingUKB_WGS_Official_release.

You will find the following scripts to perform the phasing on the UK Biobank WGS data:

  • step01_qc_per_chunks.sh : perform a quality control on each WGS chunk.
  • step02_concat_qc.sh : concat all QCeed chunks to obtain a single file per chromosome.
  • step1_phase_common.sh : phase common variants (MAF>0.001) using large chunks.
  • step2_ligate_common.sh : ligate the chunks from the common variants phasing.
  • step3_phase_rare.sh : phase rare variants (MAF>0.001) onto the common variants scaffold using small chunks.
  • step4_concat_rare.sh : concatenate the final phasing into a single file per chromosome.


It is important to note that a genome-wide chunking has been produce for this phasing. We used large chunks (~20-25MB) to phase common variants and smaller chunks (~4-7MB) to phase rare variants. This chunking has been optimized for phasing accuracy and computational costs. We based our chunking algorithm on the number of variant per chunk, the size of chunks in megabase and the size of chunks in centimorgan, to have similar chunk parameters along the genome. It can easyly be used on another dataset in genome built GRCh38. The chunking can be found in our github repository by navigating into the chunks/ folder.


For a more detail description of the steps, see below.



Quality control

We perform quality control of the variant sites and filtered out SNPs and indels for (i) Hardy-Weinberg p-value < 10-30, (ii) more than 10% of the individuals having no data (GQ score=0; missing data), (iii) heterozygous excess less than 0.5 or greater than 1.5, and (iv) alternative alleles with AAscore < 0.5. Additionally, we keep only variant sites with the tag “FILTER=PASS”, as suggested by the data providers (Halldorsson et al., Nature 2022).

The following code is an example of our QC for the chromosomes 20 in 1288 chunks. This number has to be adapted per chromosome.

#!bin/bash

# Step1. perform a QC on each WGS chunk of data
for CHR in 20; do

	ODIR0=PhasingWGS_Official_release/step0_qc/chunks/support
	dx mkdir -p ${ODIR0}

	JOBID0=$(dx run app-swiss-army-knife -icmd="tabix /mnt/project/Bulk/Whole\ genome\ sequences/Whole\ genome\ GraphTyper\ joint\ call\ pVCF/QC/qc_metrics_graphtyper_v2.7.1_qc.tab.gz chr${CHR} | awk 'BEGIN{print \"##fileformat=VCFv4.2\n##contig=<ID=chr${CHR}>\n#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\"}(10^$14<1 && 10^$14>0){AF=10^\$14;MAF=(AF>0.5?1-AF:AF);ExcHet=\$17/(2*(\$16+\$17+\$18)*(AF)*(1-AF));}{if (\$7>0.5 && \$15 <15012 && ExcHet >=0.5 && ExcHet <= 1.5){ split(\$3, array, \":\"); print \$1\"\t\"\$2\"\t\"\$4\"\t\"array[3]\"\t\"array[4]\"\t.\t.\t.\t.\"}}' | bcftools view -Ob -o pass_qc_chr${CHR}.bcf && bcftools index --threads 2 -f pass_qc_chr${CHR}.bcf" --tag filter_tab --tag tabix --tag swiss_army_knife --instance-type mem1_ssd1_v2_x2 --folder="${ODIR0}" --name qc_step03a --priority normal -y | tail -n1 | cut -d" " -f3)

	#the number of chunks is manual but can easily be stored in an array using dx find data for the other chromosomes.
	ODIR1=PhasingWGS_Official_release/step0_qc/chunks
	dx mkdir -p ${ODIR1}
	for CHUNK in {0..1288}; do
		START=$(((CHUNK*50000)+1))
		END=$(((CHUNK+1)*50000))
		JOBID1=$(dx run app-swiss-army-knife -iin="/Bulk/Whole\ genome\ sequences/Whole\ genome\ GraphTyper\ joint\ call\ pVCF/ukb23352_c${CHR}_b${CHUNK}_v1.vcf.gz" -iin="/Bulk/Whole\ genome\ sequences/Whole\ genome\ GraphTyper\ joint\ call\ pVCF/ukb23352_c${CHR}_b${CHUNK}_v1.vcf.gz.tbi" -iin="${ODIR0}/pass_qc_chr${CHR}.bcf" -iin="${ODIR0}/pass_qc_chr${CHR}.bcf.csi" -icmd="bcftools annotate -x ^INFO/AC,^INFO/AN,^FORMAT/GT -Ou ukb23352_c${CHR}_b${CHUNK}_v1.vcf.gz | bcftools view -f PASS -Ou | bcftools norm -m -any -Ou | bcftools view -i 'ALT!=\"*\"' -Ob -o split_ukb23352_c${CHR}_v1.bcf && bcftools index -f split_ukb23352_c${CHR}_v1.bcf && bcftools isec -c none -n=2 -w1 -r chr${CHR}:${START}-${END} split_ukb23352_c${CHR}_v1.bcf pass_qc_chr${CHR}.bcf -Ob -o pass_qc_ukb23352_c${CHR}_${START}_${END}_v1.bcf && bcftools index -f pass_qc_ukb23352_c${CHR}_${START}_${END}_v1.bcf && bcftools +fill-tags -r chr${CHR}:${START}-${END} -Ou pass_qc_ukb23352_c${CHR}_${START}_${END}_v1.bcf -- -t HWE -S /mnt/project/data/ukb_wgs/unphased/qc/support/UKB_samples_with_WGS_british_single_gbr.txt | bcftools view -G -e \"INFO/HWE_GBR < 1e-30\" -Ob -o pass_hwe_chr${CHR}_${START}_${END}.bcf && bcftools index -f pass_hwe_chr${CHR}_${START}_${END}.bcf && bcftools isec -c none -n=2 -w1 -r chr${CHR}:${START}-${END} pass_qc_ukb23352_c${CHR}_${START}_${END}_v1.bcf pass_hwe_chr${CHR}_${START}_${END}.bcf -Ob -o ukb23352_c${CHR}_${START}_${END}_v1.bcf && bcftools index -f ukb23352_c${CHR}_${START}_${END}_v1.bcf && rm -f pass* split_*" --tag qc --tag filter_tab --tag filter_hwe --instance-type mem1_ssd1_v2_x2 --folder="${ODIR1}" --depends-on ${JOBID0} --name qc_chr${CHR}_chunk${CHUNK} --priority normal -y | tail -n1 | cut -d" " -f3)
		
		
	done
done


#!bin/bash


# Step2. Concat all QCeed chunks into a single file per chromosome.

#!bin/bash

for CHR in 20; do
	ODIR1=PhasingWGS_Official_release/step0_qc/chunks
	ODIR2=PhasingWGS_Official_release/step0_qc/support
	ODIR3=PhasingWGS_Official_release/step0_qc
	dx mkdir -p ${ODIR2}
	dx mkdir -p ${ODIR3}
	
	dx find data --folder "${ODIR1}" --name "*c${CHR}*.bcf" --delim | sort -k 4 -t$'\t' -V | awk '{print "/mnt/project"$6}' > concat_chr${CHR}.txt
	dx upload concat_chr${CHR}.txt --path="${ODIR2}";
	JOBID0=$(dx run app-swiss-army-knife -icmd="bcftools concat -f /mnt/project/${ODIR2}/concat_chr${CHR}.txt -n -o ukb23352_c${CHR}_qc_v1.bcf --threads 2 && bcftools index ukb23352_c${CHR}_qc_v1.bcf --threads 2 -f" --instance-type mem1_ssd1_v2_x2 --folder="${ODIR3}" --name concat_qc --priority normal -y | tail -n1 | cut -d" " -f3)
	JOBID1=$(dx run app-swiss-army-knife -icmd="bcftools view -G /mnt/project/${ODIR3}/ukb23352_c${CHR}_qc_v1.bcf --threads 4 -Ob -o ukb23352_c${CHR}_qc_v1_sites.bcf && bcftools index ukb23352_c${CHR}_qc_v1_sites.bcf -f --threads 4" --tag sites --instance-type mem1_ssd1_v2_x4 --folder="${ODIR3}" --name get_sites --priority normal --depends-on ${JOBID0} -y | tail -n1 | cut -d" " -f3)

done


Phasing

SHAPEIT5 phases common variants using the SHAPEIT5_phase_common tool. As an input, SHAPEIT5_phase_common requires an unphased file (with AC and AN tags), and automatically sub-sets the file to the desired MAF (e.g. 0.001). There are different strategies to phase common variants. The first, is to phase the whole chromosome in a single job. This is feasible for SNP array data and WES data, but it is not optimal for WGS data. Therefore we recommend to chunk the chromosome into large chunks (e.g. 20 cM) if using large WGS data. In the following we see how to phase WGS data in chunks.

Phasing common variants in chunks

When using WGS data on large sample size, it is good practice to run SHAPEIT5_phase_common in different large regions of the chromosomes (e.g. 20cM). In the following, we perform phasing in chunks with overlapping regions that are large enough to have a good amount of heterozygous sites for the ligation step (i.e, assembling all chunks together).

The chunks can be found here (link available soon).

IMPORTANT: in the following code make sure to change the shapeit5 docker image name (here shapeit5_beta.tar.gz) to the latest version that you’ve downloaded here

#!/bin/bash

# step0. Create a docker folder of the RAP and upload the provided shapeit5 docker image

dx mkdir -p docker/
dx upload ../../docker/shapeit5_v1.0.0.tar.gz --path="docker/"
BIN=docker/shapeit5_v1.0.0.tar.gz

# step1. Create output directory.
ODIR=PhasingWGS_Official_release/step1_phase_common/chunks
dx mkdir -p ${ODIR}

# step2. Upload map files
cp ../../maps/genetic_maps.b38.tar.gz ./
tar xvzf genetic_maps.b38.tar.gz
dx mkdir -p data/shapeit_maps/
dx upload *.b38.gmap.gz --path="data/shapeit_maps/"


# step3. Get phasing chunks
cp ../../chunks/chunks.b38.tar.gz ./
tar -xvzf chunks.b38.tar.gz


# step4. phasing common variants (MAF>0.001)
for CHR in {1..22}; do

	CHUNKS=chunks/phase_common/chunks_chr${CHR}.txt
	BCF=/mnt/project/PhasingWGS_Official_release/step0_qc/ukb23352_c${CHR}_qc_v1.bcf # QCeed data from step0, unphased, full chromosome, full sample set.
	MAP=/mnt/project/data/shapeit_maps/chr${CHR}.b38.gmap.gz

	while read LINE; do
		REG=$(echo $LINE | awk '{ print $3; }')
		CHUNK_NBR=$(echo $LINE | awk '{ print $1; }')
		OUT=UKB_chr${CHR}.chunk_${CHUNK_NBR}.shapeit5_common.bcf
		LOG=UKB_chr${CHR}.chunk_${CHUNK_NBR}.shapeit5_common.log
		TIM=UKB_chr${CHR}.chunk_${CHUNK_NBR}.shapeit5_common.time
		
		dx run app-swiss-army-knife -iimage_file="/${BIN}" --folder="${ODIR}/" -icmd="/usr/bin/time -vo $TIM SHAPEIT5_phase_common_static_v1.0.0 --input $BCF --map $MAP --output $OUT --thread 72 --log $LOG --filter-maf 0.001 --region $REG && bcftools index -f $OUT --threads 72" --instance-type mem1_ssd1_v2_x72 --priority normal --name WGS_shapeit5_common_chr${CHR}_${CHUNK_NBR} -y
		
	done < ${CHUNKS}
done



The full list of options for the SHAPEIT5_phase_common command can be found here.

One advantage of this approach is that these two chunks can run in parallel and they require less computational resources than the single job. However, as we wish to perform phasing at rare variants using a haplotype scaffold, we need to ligate these chunks to create a single file for the entire region.


Ligate chunks

Ligation of the phased common variants is performed using the SHAPEIT5_ligate tool. The program requires an ordered list of the phased common variants. We recommend using appropriate naming of the files in the previous step, so that a command such as ls -1v can directly produce the list of files in the right order.

#!bin/bash

BIN=docker/shapeit5_v1.0.0.tar.gz
ODIR=PhasingWGS_Official_release/step1_phase_common/

threads=36
for CHR in {1..22}; do
	dx run app-swiss-army-knife -iimage_file="/${BIN}" --folder="${ODIR}/" -icmd="ls -1v /mnt/project/PhasingWGS_Official_release/step1_phase_common/chunks/UKB_chr${CHR}.chunk_*.shapeit5_common.bcf > list_ligate.chr${CHR}.txt && SHAPEIT5_ligate_static_v1.0.0 --input list_ligate.chr${CHR}.txt --output UKB_chr${CHR}.shapeit5_common_ligate.bcf --thread ${threads} --index" --instance-type mem1_ssd1_v2_x8 --priority low --name WGS_shapeit5_chr${CHR}_ligate -y

done

At the end of this step, we have a region-wide haplotype scaffold.


Phasing rare variants in chunks

Having obtained the haplotype scaffold at common variants in the previous step, we can now phase rare variants using the SHAPEIT5_phase_rare tool. We always do this in chunks, in order to maximise the parallelization of our jobs (e.g. in 5Mb chunks). The program requires two input files, one is the haplotype scaffold generated in the previous steps, and the other input required is a file containing the whole unphased region, the same input file provided for SHAPEIT5_phase_common: SHAPEIT5 automatically discards common variants for this file, duplicated in the phased scaffold.

#!/bin/bash


BIN=docker/shapeit5_v1.0.0.tar.gz
threads=36

# step0. Create output directory.
ODIR=PhasingWGS_Official_release/step2_phase_rare/chunks
dx mkdir -p ${ODIR}

# step1. phasing rare variants (MAF<0.001)
for CHR in {1..22}; do
	CHUNKS=chunks/phase_rare/chunks_chr${CHR}.txt
	BCF=/mnt/project/PhasingWGS_Official_release/step0_qc/ukb23352_c${CHR}_qc_v1.bcf # QCeed data from step0, unphased, full chromosome, full sample set.
	MAP=/mnt/project/data/shapeit_maps/chr${CHR}.b38.gmap.gz
	SCAF=/mnt/project/PhasingWGS_Official_release/step1_phase_common/UKB_chr${CHR}.shapeit5_common_ligate.bcf	
	
	while read LINE; do

		CHUNK_NBR=$(echo $LINE | awk '{ print $1; }')
		
		SCAFFOLD_REG=$(echo $LINE | awk '{ print $3; }')
		SCAFFOLD_REG_START=$(echo ${SCAFFOLD_REG} | cut -d":" -f 2 | cut -d"-" -f1)
		SCAFFOLD_REG_END=$(echo ${SCAFFOLD_REG} | cut -d":" -f 2 | cut -d"-" -f2)
		SCAFFOLD_REG_NAME=${CHR}_${SCAFFOLD_REG_START}_${SCAFFOLD_REG_END}
			
		INPUT_REG=$(echo $LINE | awk '{ print $4; }')
		INPUT_REG_START=$(echo ${INPUT_REG} | cut -d":" -f 2 | cut -d"-" -f1)
		INPUT_REG_END=$(echo ${INPUT_REG} | cut -d":" -f 2 | cut -d"-" -f2)
		INPUT_REG_NAME=${CHR}_${INPUT_REG_START}_${INPUT_REG_END}

		OUT=UKB_chr${CHR}.chunk_${CHUNK_NBR}.shapeit5_rare.bcf
		LOG=UKB_chr${CHR}.chunk_${CHUNK_NBR}.shapeit5_rare.log
		TIM=UKB_chr${CHR}.chunk_${CHUNK_NBR}.shapeit5_rare.time
		
		dx run app-swiss-army-knife -iimage_file="/${BIN}" --folder="${ODIR}/" -icmd="/usr/bin/time -vo $TIM SHAPEIT5_phase_rare_static_v1.0.0 --input-plain $BCF --map $MAP --output $OUT --thread ${threads} --log $LOG --scaffold $SCAF --scaffold-region $SCAFFOLD_REG --input-region $INPUT_REG && bcftools index -f $OUT --threads ${threads}" --instance-type mem3_ssd1_v2_x48 --priority normal --name WGS_shapeit5_rare_chr${CHR}_${CHUNK_NBR} -y
		
	done < ${CHUNKS}
done

The full list of options for the SHAPEIT5_phase_rare command can be found here.


Concatenate chunks

Since the phasing of rare variants has been performed using a ligated haplotype scaffold, the chunks can now be concatenated together to resolve the phasing for the entire chromosome. This is done using the bcftools concat command.

#!/bin/bash

threads=16

for CHR in {1..22}; do
	ODIR=PhasingWGS_Official_release/step2_phase_rare/concat/chr${CHR}
	dx mkdir -p ${ODIR}	
	chunks=/mnt/project/PhasingWGS_Official_release/step2_phase_rare/chunks/UKB_chr${CHR}.chunk_*.shapeit5_rare.bcf
	OUT=UKB_chr${CHR}.full.shapeit5_rare.bcf
	
	dx run app-swiss-army-knife --folder="${ODIR}/" -icmd="ls -1v ${chunks} > concat_list_chr${CHR}.txt && bcftools concat -n -f concat_list_chr${CHR}.txt -o ${OUT} && bcftools index ${OUT} --threads ${threads} && rm concat_list_chr${CHR}.txt" --instance-type mem1_ssd1_v2_x16 --priority normal --name WGS_concat_chr${CHR} -y
	
done


Validation of your phasing

You can validate the quality of the haplotypes using the SHAPEIT5_switch tool. For this you will need parent-offspring duos or trios stored in a three-columns file (here called family.ped) following this format:

  • family.ped: offspring_id parent1_id parent2_id

To validate the phasing using family data, the phasing must be performed by excluding parental genomes, so that offsprings are phased regardless of their parental genomes. This can be done using the bcftools view command, with as input the output data of the quality control step (located here if you used our pipeline /mnt/project/PhasingWGS_Official_release/step0_qc).

ODIR=PhasingWGS_Official_release/benchmark
dx mkdir -p ${ODIR}
for CHR in {1..22}; do
	IN=/mnt/project/PhasingWGS_Official_release/step0_qc/ukb23352_c${CHR}_qc_v1.bcf
	OUT=benchmark_ukb23352_c${CHR}_qc_v1.bcf
	dx run app-swiss-army-knife --folder "${ODIR}" -icmd="bcftools view --threads 16 -S ^parents.txt -Ob -o ${OUT} ${IN} && bcftools index ${OUT} --threads 16" --instance-type mem1_ssd1_v2_x16 --priority normal --name phasing_chr${CHR} -y
done

After excluding parental genomes using the above command, proceed with the normal phasing procedure as detailed above.

Let’s consider that you performed the above steps of phasing using the input data exluding parental genomes, which produced a phased output file that you named benchmark_ukb23352_c${CHR}_qc_v1.phased.bcf). You can validate you phasing using the following command:

BIN=docker/shapeit5_v1.0.0.tar.gz
ODIR=PhasingWGS_Official_release/benchmark/
IDIR=/mnt/project/PhasingWGS_Official_release/step0_qc

VAL=${IDIR}/ukb23352_c${CHR}_qc_v1.bcf # This file is the QCeed file (from step0) that contains all individuals, including parental genome.
EST=/mnt/project/PhasingWGS_Official_release/benchmark/benchmark_ukb23352_c${CHR}_qc_v1.phased.bcf # This file is the phased WGS from which parental genome have been excluded.

dx mkdir -p ${ODIR}
for CHR in 20; do
	dx run app-swiss-army-knife --folder "${ODIR}" -iimage_file="/${BIN}" -icmd="SHAPEIT5_switch_v1.0.0 --validation ${VAL} --estimation ${EST} --region chr${CHR} --output benchmark_chr${CHR}"  --instance-type mem2_ssd1_v2_x16 --priority normal --name benchmark_chr${CHR} -y
done

The full list of options for the SHAPEIT5_switch command can be found here.


Back to top

Copyright © 2022-2023 Olivier Delaneau | All Rights Reserved | SHAPEIT5 executables and source code are distributed under the MIT license.