Skip to main content

Tutorial - Analyzing a phased trio

Introduction​

In some scenarios, it can be useful to gather genome sequencing data from both the parents and offspring...this can form a "trio" (e.g. mom, dad, and child)

After doing variant calling and producing a VCF file, the VCF data can additionally be "phased" which tells us information about each "haplotype" of a VCF dataset

For this tutorial, we will not go through the process of creating such a file, but will look at trio from the 1000 genomes dataset for a Kinh-Vietnamese trio (KHV in filename)

You can either use the JBrowse CLI or GUI to add this file to JBrowse. Note that this file only represents chr1

After adding the VCF it will look like this

Initial load of the VCF file, showing the default display mode with simple orange boxes for each variant
Initial load of the VCF file, showing the default display mode with simple orange boxes for each variant

Enabling the matrix view​

JBrowse can show variant datasets using a specialized "display mode" called the "Multi-sample variant display (matrix)"

This mode produces a new visualization modality that shows all the samples at once, along with all the variants, in a dense "heatmap" or "matrix" style display

Navigating to the track menu, and selecting 'Multi-sample variant display (matrix) enables all the samples to be given their own row in the view. Each variant is a 'column' in this matrix, and each sample is a 'row'. Black lines connect the variants to their genome position
Navigating to the track menu, and selecting 'Multi-sample variant display (matrix) enables all the samples to be given their own row in the view. Each variant is a 'column' in this matrix, and each sample is a 'row'. Black lines connect the variants to their genome position

In the above figure, the default coloring for this matrix view is enabled.

We won't be focusing on this here, see rice example for more info on this coloring

Enabling the "phased mode" of the matrix view​

The matrix mode offers several options, one of which is the "phased" mode. If you have phased genotypes (e.g. you have genotypes with a vertical bar 0|1) for at least some of your variants, then this rendering mode will be available

The ideal is that your variants will be "completely phased". This sometimes requires specialized programs like SHAPEIT

Screenshot showing the phased rendering mode along with the menu item used to select it 'Rendering mode'->'Phased'
Screenshot showing the phased rendering mode along with the menu item used to select it 'Rendering mode'->'Phased'

Finding matching haplotypes with "visual phasing"​

The term "visual phasing" comes from the genetic genealogy subfield. I am borrowing it here, but the idea is simple: you can look at the genotype matrix here, and see areas where different rows are matching. You would expect that the child would match the mom in some places, and the dad in the other. And indeed, each row looks somewhat like a barcode, so you can find matching pieces like this

Screenshot showing the phased rendering mode without any added markup. You can look at this figure and see various areas where rows match one another. The first two rows are the two haplotypes of the child, next two rows are the two haplotypes of the mom, and next two rows are the two haplotypes of the father
Screenshot showing the phased rendering mode without any added markup. You can look at this figure and see various areas where rows match one another. The first two rows are the two haplotypes of the child, next two rows are the two haplotypes of the mom, and next two rows are the two haplotypes of the father

Using a program to help find phased blocks​

In the above section, we could see matching blocks with our eyes, but can a program help?

There are indeed options for this. The program hap-ibd (https://github.com/browning-lab/hap-ibd) can find "identical by descent" blocks. It is actually capable of finding such blocks in massive population-scale datasets, but we are applying it here to a smaller trio VCF. The hap-ibd program takes as input:

Background: relationship between phased blocks, and the biology of recombination​

Many people have heard of the term "recombination" or "crossing over" with regards to DNA, but what is it?

NHGRI definition: "Crossing over, as related to genetics and genomics, refers to the exchange of DNA between paired homologous chromosomes (one from each parent) that occurs during the development of egg and sperm cells (meiosis)." (https://www.genome.gov/genetics-glossary/Crossing-Over)

But this is an important point: Your parents genomes don't recombine during fertilization. Instead, your grandparents genomes recombine during the production of the parents gametes.

Converting hap-ibd data into a format for JBrowse​

By loading the hap-ibd data in JBrowse, hap-ibd can tell us which blocks of the chromosome are matches

#!/bin/bash
# create a bed file from hap-ibd output with columns "chr, start, end, sample_name1 hap1 sample_name2 hap2"
# add this config to jbrowse get(feature,'sample1')+':HP'+get(feature,'hap1')+'     '+get(feature,'sample2')+':HP'+get(feature,'hap2')
zcat result.ibd.gz |  cut -f 5,6,7 >! out.bed
zcat result.ibd.gz |  cut -f 1,2,3,4 >! samples.txt
echo "#chr	start	end	sample1	hap1	sample2	hap2" > out.bed
paste out.bed samples.txt >> out.bed

After this conversion, we can load this simple BED file into JBrowse via the GUI or the CLI. It is probably small enough that it doesn't even need tabix conversion

Visualizing phased blocks and crossing over points in phased VCF files in JBrowse​

After loading the hap-ibd track, we can see the blocks that hap-ibd calculated. We can additionally connect the lines onto the matrix view (which admittedly isn't straightforward, we have to follow the lines from the genomic position to the matrix position),

We can see this in the trio dataset where the child has a mixture of the dad's haplotypes, and a mixture of mom's haplotypes

Screenshot showing the connection between hap-ibd annotations (orange) and the phased VCF matrix view. The colored blocks are marked up using Google Slides. As a result of this visualization, we can see a crossing-over point that occurred (independently) in both the mom and dad at almost the same position, which form continuous blocks in the child.
Screenshot showing the connection between hap-ibd annotations (orange) and the phased VCF matrix view. The colored blocks are marked up using Google Slides. As a result of this visualization, we can see a crossing-over point that occurred (independently) in both the mom and dad at almost the same position, which form continuous blocks in the child.

In the above screenshot, you can look at the 'barcode-like' patterns to see the matches between MOM A1 (allele 1) in mom and the child, MOM A2 (allele 2) in mom and child, DAD A1 (allele 1) in dad and child, DAD A2 (allele 2) in dad and child

Footnote​

The above procedure, particularly the final visualization with marked up blocks, was done manually. It may be able to be improved in the future!

Author note: I found it quite wonderful to see this type of visualization! Feel free to let me know if I missed anything or if you have any other feedback

Live link https://jbrowse.org/code/jb2/main/?config=%2Fgenomes%2FGRCh38%2F1000genomes%2Fconfig_1000genomes.json&session=share-4gbEzsiqFe&password=Q2O2L