A reference genome is a selected representation of an organism’s genome used as a common framework for locating biological features and comparing genetic sequences. It usually consists of assembled DNA sequences organized into chromosomes or smaller sequence units, with identifiers and coordinates that allow researchers to describe where a sequence or feature occurs. A reference may derive from one individual, several individuals, or a collection of representative sequences. It is not necessarily complete, genetically typical, or representative of every member of a species.
What a reference represents
A reference genome serves two related purposes: it is a sequence against which other sequences can be compared, and it is a coordinate system for recording biological information. A gene, for example, can be described by its chromosome, start and end positions, and orientation relative to a particular reference assembly.
The word reference describes the sequence’s role rather than guaranteeing a particular level of quality. An incomplete draft can serve as a reference, while a highly accurate assembly may remain a sample-specific resource rather than a community standard.
Several distinctions are important:
- Reference versus individual genome. A reference is a chosen framework; an individual genome is the genetic material of a particular organism. A reference assembled from multiple donors need not correspond exactly to any individual.
- Reference versus consensus. A consensus sequence selects a representative base at each aligned position according to a specified rule. A reference need not contain the most frequent base at every position.
- Reference versus ancestral sequence. The base present in a reference is not necessarily the ancestral state from which other variants evolved.
- Reference versus “normal.” A reference allele is simply the allele represented in the reference sequence. That designation alone says nothing about its frequency, biological effect, or association with disease.
Many traditional references are primarily haploid representations: they provide one sequence path for each chromosome type rather than separately representing both inherited chromosome copies. Some also include alternative sequences for regions with substantial variation.
Assembly and annotation
Constructing an assembly
A genome assembly is produced by reconstructing longer sequences from the fragments generated by DNA sequencing. Overlapping or otherwise connected sequences are combined into contigs, continuous stretches without assembly gaps. Contigs may then be ordered and oriented into scaffolds. Additional evidence can place scaffolds along chromosomes.
Assembly can draw on short, highly accurate reads; long reads that span repetitive regions; and long-range information about sequence order or chromosome organization. Assembly without using an existing reference as the primary template is called de novo assembly.
Difficult regions include long repeats, closely related duplicated sequences, centromeres, and sequences present in multiple copies. These regions can be collapsed, incorrectly joined, or left unresolved. In genomes with substantial differences between chromosome copies, an assembly can also mix the two parental sequences or mistakenly represent an alternative copy as an additional region.
Some modern assemblies are haplotype-resolved: they separate sequences into distinct haplotypes, often corresponding to the chromosome sets inherited from each parent. This preserves variation that a single sequence path cannot fully represent.
Sequence components
A reference assembly may contain:
- Chromosome sequences, representing assembled chromosomes.
- Unlocalized scaffolds, assigned to a chromosome but not to a precise position.
- Unplaced scaffolds, not confidently assigned to a chromosome.
- Alternative loci, providing additional representations of variable regions.
- Patch sequences, supplying corrections or alternative representations without necessarily changing the established chromosome coordinates.
Reference packages may also include organellar sequences, such as the genome of a mitochondrion. The exact sequence collection matters: two files described informally as the same reference can differ in their inclusion of alternatives, patches, or supplementary sequences.
Annotation
Genome annotation identifies features such as genes, transcripts, regulatory regions, and repetitive elements on an assembly. Assembly and annotation are distinct products. The assembly supplies the underlying sequence and coordinates; annotation assigns biological interpretations to intervals within that sequence.
Annotation may incorporate experimental evidence from RNA sequencing, known proteins, and comparisons with other organisms. An annotation release can change even when the reference sequence remains unchanged. Reproducible analysis therefore requires both the assembly version and, where relevant, the annotation version.
Use in genomic analysis
Read alignment and variant detection
A common workflow aligns sequencing reads to a reference using sequence alignment. The alignment estimates where each read originated and records matches, mismatches, insertions, and deletions.
Variant calling then evaluates the evidence for differences between the sample and the reference. These differences can include single-base substitutions, short insertions or deletions, and larger structural variants, such as inversions, duplications, and rearrangements.
A reference makes comparison computationally practical and allows results from different samples to share a common coordinate system. Nevertheless, a difference from the reference is not automatically a newly arisen mutation or a biologically important change. Its interpretation depends on inheritance, population frequency, experimental evidence, and biological context.
Functional and comparative analysis
Reference genomes support the measurement of gene expression by providing locations to which transcript-derived reads can be assigned. They also underpin studies of DNA–protein interactions, epigenetic modifications, and other molecular features.
In genome-wide association studies, shared genomic coordinates allow genetic markers and statistical associations to be compared across datasets. In population genetics, references help organize variation across populations, although the reference’s composition can influence which variation is detected.
References also support comparative genomics, including the identification of related genes, conserved regions, and changes in chromosome organization. In bacteria and other microorganisms, a strain-specific reference may facilitate comparisons among closely related samples, but it cannot capture all genes present across the species.
Coordinates, versions, and reproducibility
Genomic coordinates have meaning only in relation to a specified assembly. If an assembly is corrected, extended, or rearranged, the same biological sequence may acquire different coordinates. A chromosome position without an assembly identifier is therefore potentially ambiguous.
Different file formats also use different conventions. Some count the first base as position 1 and include both interval endpoints. Others use zero-based, half-open intervals, in which the start is included and the end is excluded. Mixing these conventions can shift a feature by one base or alter its reported length.
Assembly names alone may not fully identify an analysis reference. Important details include:
- The assembly accession and version.
- The exact set of sequence records included.
- Whether alternative loci, patches, or supplementary sequences were used.
- The annotation release.
- Sequence names and coordinate conventions.
Coordinate-conversion tools, often called liftover tools, map locations between assemblies using alignments. Conversion is not always one-to-one: an interval may be absent, split, duplicated, or substantially changed in the destination assembly. Converting a variant can also require checking the destination reference allele and its local sequence context.
Human reference genomes
The Human Genome Project established the foundation for widely used human reference sequences. A draft was published in 2001, and completion of the project was announced in 2003. The resulting reference was a mosaic assembled from multiple donors rather than a complete representation of one person’s two chromosome sets.
The Genome Reference Consortium subsequently maintained human assemblies including GRCh37, released in 2009, and GRCh38, released in 2013. Both became important coordinate systems for genomic datasets. Their continued use reflects the accumulated value of annotations, software, and published results tied to particular assembly versions.
In 2022, the Telomere-to-Telomere Consortium published the T2T-CHM13 assembly, providing gapless sequences for the represented human chromosomes and resolving many regions that had been absent or incomplete in earlier references. CHM13 did not contain a Y chromosome; a complete human Y-chromosome sequence from a different sample was published in 2023. These advances improved sequence completeness without making any single assembly representative of all human genetic diversity.
Limitations and reference bias
A single linear sequence cannot simultaneously represent all alternative sequences, gene-content differences, and chromosome arrangements within a species. This creates several limitations.
Mapping bias. Reads resembling the reference often align more readily than reads containing divergent alleles or large sequence differences. This can distort allele measurements and reduce sensitivity to non-reference variation.
Missing sequence. DNA absent from the reference has no direct coordinate within its main sequence. Reads originating from such regions may fail to align or may align incorrectly elsewhere.
Ambiguous placement. Repeated or nearly identical sequences can make it difficult to determine a read’s true origin, even when the relevant sequence is present.
Incomplete representation of diversity. A reference derived from a limited set of samples may omit variation found in other populations, breeds, strains, or lineages.
Assembly errors. Collapsed duplications, incorrect joins, and residual base errors can be mistaken for biological differences or obscure real ones.
These limitations depend on the reference, sequencing technology, analysis method, and biological question. Reference-based and reference-free approaches therefore provide complementary information.
Pangenomes and graph references
A pangenome represents genomic content across multiple individuals or lineages rather than relying on one sequence alone. In microorganisms, the term often describes the combined gene repertoire of a group, including shared and variably present genes. In human genomics, pangenome resources can include multiple assembled haplotypes and their sequence relationships.
A graph reference represents alternative sequences as paths through a graph. Shared regions can use common paths, while substitutions, insertions, deletions, and more complex differences create alternatives. A draft human pangenome reference published in 2023 demonstrated this multi-assembly approach on a substantially broader basis than a single linear reference.
Graph and multi-reference methods can reduce some forms of reference bias and improve access to previously poorly represented variation. They also create challenges for indexing, alignment, annotation, and stable coordinate naming. Linear references remain widely used because their coordinates and associated tools are deeply established, while pangenome approaches provide a more explicit representation of genomic diversity.