Lesson 3 of 13 · 11 min

Reference Genomes and Genome Builds

A reference genome is a single, standardized DNA sequence that scientists agree to use as a shared map for a species. It is a composite (sometimes called a mosaic), meaning it was assembled from the DNA of several donors and stitched into one representative sequence rather than copying any single person. A sequence here just means the ordered string of the letters A, C, G, and T that spell out DNA.

Almost every genomics analysis compares your own data against this reference. When you sequence a sample, the short pieces of DNA you read are lined up against the reference so that each position can be named by a chromosome and a coordinate. The places where your sample differs from the reference are what we later call variants.

Human genome builds

A build, also called an assembly, is one finished version of the reference. For humans the two builds you will meet are GRCh37 and GRCh38, released by the Genome Reference Consortium, where GRCh stands for Genome Reference Consortium human. UCSC gives the same two builds shorter nicknames, so GRCh37 is hg19 and GRCh38 is hg38, and hg simply means human genome.

Tip: Builds also receive patches, written like GRCh38.p14, which add small corrections without moving any existing coordinates. A brand-new build such as GRCh38 does shift coordinates, so the same position number in GRCh37 usually points to a different base in GRCh38. Never mix builds in one analysis; convert coordinates first with a tool like UCSC liftOver or CrossMap.

Chromosome naming

The same chromosome can be written in different ways depending on the source. Ensembl names it 1, UCSC writes chr1 with a chr prefix, and NCBI often uses an accession such as NC_000001.11. These names all label the identical sequence, but software compares them as plain text, so chr1 and 1 look like two different chromosomes and can make a pipeline silently fail.

Where genomes are hosted

Three public archives host the human reference and its builds. Ensembl at ensembl.org, NCBI at ncbi.nlm.nih.gov, and UCSC at genome.ucsc.edu all serve the same underlying assembly but with their own file layouts and naming styles. Pick one source and one build at the start of a project and stay consistent throughout.

bash

# Download chromosome 21 of the GRCh38 build from Ensembl (a gzip-compressed FASTA file).
# curl fetches a file over the internet; the -O flag saves it under its original remote name.
curl -O http://ftp.ensembl.org/pub/release-110/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.chromosome.21.fa.gz

# Uncompress the .gz file into a plain-text .fa (FASTA) file.
gunzip Homo_sapiens.GRCh38.dna.chromosome.21.fa.gz

# Print only the header line. In grep, "^>" means "lines that start with the > character",
# which is how every FASTA sequence names itself.
grep "^>" Homo_sapiens.GRCh38.dna.chromosome.21.fa

>21 dna:chromosome chromosome:GRCh38:21:1:46709983:1 REF

Try it yourself: Open the Ensembl or UCSC website and locate the GRCh38 assembly for one chromosome. On Ensembl, go to ensembl.org, search for Human, then open Chromosome 21; on UCSC, go to genome.ucsc.edu, launch the Genome Browser, choose the Human GRCh38/hg38 assembly, and enter chr21. Notice that Ensembl labels it 21 while UCSC labels it chr21, even though both display the exact same DNA.

Genes, Chromosomes, and Non-coding DNA From Sanger to Next-Generation Sequencing