Variant Records and Allele Normalization

Why Variant Representation Matters

A small genomic change can be described in more than one way. This is especially common for insertions and deletions in repeated sequence: deleting one of several consecutive identical bases can produce the same resulting DNA sequence, even when the stated position differs. Before comparing variants across samples or data sets, it is therefore important to distinguish a biological sequence change from the particular record used to describe it.

This tutorial defines the basic parts of a small-variant record and introduces normalization: a set of representation rules used to make equivalent records consistent. The focus is on substitutions and short insertions and deletions, not on assigning genotypes to individuals.

Reference and Alternate Alleles

A variant description depends on a reference genome: a specified genome assembly and its sequence. For the examples here, positions are numbered from 1 within a chromosome, and a record’s position is the coordinate of the first base in its reference allele. The reference allele, written REF, is the sequence beginning at that position; an alternate allele, written ALT, replaces that sequence. The unchanged sequence before and after the record is retained.

For example, suppose position 104 in a hypothetical reference has the base C. A record with position 104, REF C, and ALT T describes replacing C with T. This is a single-nucleotide substitution. The record says that T is an alternative sequence at that locus; it does not say how common T is, whether it is associated with a trait, or which individuals carry it.

For a deletion, REF is longer than ALT; for an insertion, ALT is longer than REF. In common anchored representations, both allele strings contain at least one base, even when the biological change is an insertion or deletion. That retained base provides sequence context for locating the change. The exact representation convention matters when comparing records, so the reference genome and coordinate convention must be held fixed.

Biallelic and Multiallelic Records

The number of alternate alleles listed in a record determines whether the record is biallelic or multiallelic. “Biallelic” counts the reference allele as one of the possible alleles: a biallelic record has one REF allele and one ALT allele. A multiallelic record has one REF allele and at least two distinct ALT alleles. These terms describe the alleles listed at a locus, not the number of individuals observed or the number of copies carried by any one individual.

Normalization: A Consistent Representation

Normalization puts a variant record into a standard form so that equivalent descriptions can be compared. For small variants, two useful operations are trimming redundant shared bases and, for indels in repeated sequence, left alignment. A minimal representation removes shared sequence from the allele strings when doing so leaves valid nonempty strings. Left alignment chooses the leftmost equivalent placement on the same reference sequence.

A basic check for equivalence is to compare the alternate sequence produced in the same reference context. If the unchanged sequence to the left is \(L\), the REF string is \(R\), the ALT string is \(A\), and the unchanged sequence to the right is \(S\), the alternate sequence is \(L A S\). Two records are equivalent in that context if applying either one yields the same alternate sequence. A difference in written position alone does not establish that two indel records describe different changes.

Equivalent Insertions in a Repeat

When several neighboring reference bases are identical, the same inserted base can be placed at more than one point in the run without changing the resulting sequence. A left-aligned representation resolves this ambiguity by choosing the leftmost placement. In an anchored record, the base immediately before the insertion is retained in both allele strings.

Equivalent Deletions in a Repeat

The same representational ambiguity occurs for deletions. In an anchored representation, the base before the deleted sequence is retained in both alleles. If a run contains identical bases, deleting any one of them may yield the same sequence. Left alignment again selects the leftmost equivalent representation.

Why Normalization Matters

Without normalization, equivalent indels can appear at different coordinates or have different allele strings. A direct comparison might then treat them as distinct records, even though applying each record to the same reference produces the same alternate sequence. Normalization makes record matching and downstream summaries more reliable; it does not determine whether a variant is real, harmful, or carried by a particular person.

The reference assembly is essential. Coordinates and reference sequences can differ between assemblies, so a position-and-allele description without its reference context may be ambiguous. Normalization also has a limit: it chooses a consistent representation under a stated convention, but it cannot make records comparable if they use incompatible reference sequences or inconsistent coordinate systems.

Common Mistake

A common error is to conclude that records at different positions must describe different indels. That reasoning ignores repeated sequence. For example, deleting one C from a run of three Cs can be written at different positions, yet all such descriptions can yield the same alternate sequence. The correction is to compare the resulting sequences in the same reference context, then use minimal, left-aligned representation to make the description consistent.

A second error is to interpret “multiallelic” as meaning that one individual carries multiple alternate alleles. The term describes the alternatives listed at a locus. Whether an individual carries a particular allele is a separate question about genotype data.