Structural-Variant Classes And Event Representation

Why Structural-Variant Representation Matters

A structural variant can change a chromosome by removing, adding, reversing, or relocating sequence. Describing a call only as “a large variant” is not enough for statistical analysis: a dataset needs to say what event is being represented, where it occurs, how large it is, and which samples carry its alternate allele. These details make an event interpretable and help prevent records with different meanings from being treated as equivalent.

The same event also has more than one useful description. Its biological class describes the sequence change; its coordinates describe the affected reference sequence or the breakpoints; and its sample-level genotype describes the alleles observed in an individual. These fields are related, but none can safely be substituted for the others. In particular, event class alone does not establish the resulting copy-number state, functional consequence, or association with a phenotype.

Five Common Event Classes

The event class is determined by the relationship between the alternate chromosome and the reference sequence. The following definitions describe the sequence change, not its cause or clinical effect.

• Deletion: a stretch of reference sequence is absent from the alternate chromosome. It is usually represented by the reference interval that is removed.
• Duplication: a stretch of sequence is present in an additional copy relative to the reference. The duplicated source interval is identified; if the copy is inserted elsewhere, the destination and its breakpoints may also be needed to describe the event fully.
• Inversion: a stretch of sequence is reversed in orientation relative to the reference. Its two boundaries delimit the inverted interval.
• Insertion: sequence is added at a location relative to the reference. The inserted sequence may be novel or may originate from another genomic location. An insertion is naturally located at a breakpoint between reference bases, and its length is the number of inserted bases when that sequence or length is known.
• Translocation: sequence is joined to a different genomic location, commonly involving two chromosomes or two distant regions. It is described by the participating breakpoints and their orientations, rather than by one affected interval.

These categories are useful, but real events can be more complicated than one label suggests. For example, a rearrangement can include both an inversion and a deletion, or a duplication can be part of a larger translocation. A record should describe the event at the level supported by the call; it should not force a complex event into a simple class if that loses essential structure.

Coordinates and Lengths

A coordinate is meaningful only with a specified reference assembly and coordinate convention. A common convention for an affected reference interval is a 1-based closed interval: the first base of the reference is position 1, and both the start and end positions are included. Under that convention, an interval from \(s\) through \(e\), with \(e \geq s\), contains \(e-s+1\) bases.

The added 1 is important: subtracting the endpoints alone counts the gaps between positions, not the number of included bases. For example, positions 200 through 249 contain 50 bases. This formula applies to a reference interval, such as the sequence removed in a deletion or reversed in an inversion. An insertion’s length instead refers to the inserted sequence, not to the number of reference bases spanned by its breakpoint. A translocation has no single interval length in general; the lengths of any affected or exchanged segments must be specified separately if they are known.

Some data formats use different coordinate conventions, including zero-based half-open intervals. The same event can therefore have different numerical endpoints in different formats. Do not compare start and end values until the reference assembly and coordinate convention are known. A breakpoint may also be uncertain or represented as a range; in that case, report the uncertainty rather than implying an exact base-level boundary.

Representing Insertions and Rearrangements

An insertion lies between reference bases rather than replacing a reference interval. It can be described by the reference location of the insertion breakpoint, the inserted sequence or its length if known, and the sample genotype. If the sequence is unknown but the length is estimated, the record should make that distinction clear. A breakpoint location does not itself tell the reader how many bases were added.

An inversion is more naturally represented by its two interval boundaries. A translocation instead requires a pair of genomic breakpoints: each breakpoint has a chromosome and position, and orientation information helps specify how the segments are joined. If an event involves several joins, one pair of coordinates is insufficient; represent all supported joins or describe the event as complex. Do not call the distance between two translocation breakpoints “the event length” unless the relevant exchanged or altered segment has been defined.

Genotype Is Not Event Class

For a defined biallelic event in a diploid sample, \(0/0\), \(0/1\), and \(1/1\) conventionally indicate zero, one, or two alternate alleles, respectively. The symbols 0 and 1 refer to reference and alternate alleles for that event record; they are not universal labels for every structural variant. If a dataset uses another genotype convention, it must be documented. If phase is known and relevant, phased notation such as \(0|1\) distinguishes allele order, whereas unphased \(0/1\) does not identify which parental chromosome carries the alternate allele.

A genotype is interpretable only when the corresponding event is defined consistently. A \(0/1\) call for one deletion is not interchangeable with a \(0/1\) call for a different deletion, even if the events overlap. Nor does the class “duplication” tell us a sample’s precise copy-number state. Those questions require additional information beyond the event-class and genotype fields covered here.

For association analysis, each sample’s genotype must refer to the same event definition. At minimum, a useful record keeps the event identifier, class, assembly, coordinate convention, interval or breakpoints, length and length basis, and the sample-level genotype. When boundaries or sequence are uncertain, represent that uncertainty explicitly. This keeps uncertainty in the data record rather than hiding it behind an apparently exact coordinate.

Short Derivation: Why Inclusive Intervals Add One

For a 1-based closed interval from \(s\) to \(e\), the included integer positions are \(s, s+1, \ldots, e\). Their count is the final position minus the first position plus one. For instance, positions 7, 8, 9, and 10 are four bases, and \(10-7+1=4\). Therefore, the inclusive interval length is \(e-s+1\). If instead one uses a zero-based half-open interval, the start is included and the end is excluded, so its length is \(e-s\); applying that formula to inclusive coordinates would undercount by one.

Common Mistakes and Why Representation Matters

• Subtracting endpoints without checking the convention. For a 1-based closed interval, using \(e-s\) undercounts the number of bases by one. Confirm the coordinate convention and apply the matching length formula.
• Giving an insertion an interval length based on reference coordinates. The inserted bases are not reference bases. Report the insertion breakpoint and count the inserted sequence, if known.
• Reporting one translocation coordinate or an arbitrary “length.” A translocation joins locations, so both breakpoints and their orientations are needed to describe the join. A distance between chromosomes is not a segment length.
• Treating event class as a full biological interpretation. “Duplication” identifies a sequence-change class; by itself it does not specify the exact sample copy-number state or establish a functional effect.
• Comparing coordinates without an assembly. The same numerical position can refer to different sequence in different reference assemblies. Include the assembly with every coordinate-based description.

These distinctions matter because association analyses use sample-level calls tied to event definitions. If records disagree about the reference assembly, interval convention, breakpoint, or alternate allele, an apparent difference between samples may be a representation mismatch rather than a biological difference. Clear event descriptions do not guarantee accurate calls, but they make the meaning of each call auditable and interpretable.