genomic sequence set copy number complement The notion of a 'complement' is useful as a special case of a set, where the members necessarily comprise an exhaustive collection of *all* objects that make up some well-defined set. Here, a 'copy number complement' represents 'represents the set of *all* copies of a specified sequence in a particular genome. Note that sequences can be duplicated in a set (i.e. contain more than one member representing the same sequence). In the 'copy number complement' example, each set member is a copy of this same biological sequence. The count of how many of a particular sequences are found in a genome is the sequences 'copy number'. In diploid organisms, the normal copy number for sequences at most locations is 2 (a notable exception being those on the X-chromosome where normal copy number is 1). Variations in copy number occur if this count increases due to a duplication of the gene/region, or decreases due to a deletion of a gene/region. A driving use case for representing copy number is to support associations between variation in copy number of a particular sequence, and phenotypes or diseases that can result. A 'complement' refers to an exhaustive collection of *all* objects that make up some well-defined set. Such a set may contain 0, 1, or more than one members. The notion of a complement is useful for defining many biologically-relevant sets of sequence features, such as 'copy number complements' representing the set of all copies of a particular sequence in a genome. The fact that we are counting how many copies of the same *sequence* exist in a genome here, as opposed to how many of the same *feature*, is what sets sequence-level concepts like 'copy number complement' apart from feature-level concepts like 'single locus complement'. To illustrate the difference, consider a duplication event that creates a new copy of the human APOE gene on a different chromosome. This creates an entirely new sequence feature at a distinct locus from that of the original APOE gene. The 'copy number complement' for sequence defined by the APOE gene locus would have a count of three, as this sequence is present three times in the genome. But the 'single locus complement' at the APOE gene locus would still have a count of two - because the duplicated copy is at a different location in the genome, and therefore does not represent a copy of the APOE locus. A set representing the complement of all copies of a particular biological sequence (typically at the scale of complete genes or larger) present in a particular genome. The identity of a 'copy number complement' instance is determined by the sequence defining its members, and their count (the number of times this sequence appears in a particular genome). In reality the sequence of each copy may not be identical, given the tendency of large regions to accumulate subtle variations. What matters is that they share a common origin/alignment with a defining location in a reference genome. We represent the notion of copy number at the "sequence level" (as opposed to the "sequence feature level") because we are concerned only with the number of copies of a sequence in a genome, and not the location of the features bearing this sequence. Consider a copy number complement comprised of three copies of the sequence defined by the location Chr8 100000-200000 on a GRCh38.2 reference genome. In one person's genome, this sequence may appear at its normal location on Chromosome 8, as well as in duplications on chromosomes 5, and 12. In another genome the sequence might appear three times as well, but on chromosomes 8, 9, and 15. When representing causal associations linking copy number to disease, it is important that these are considered to be *the same* copy number complement - because what a curator associates with a disease is the presence of three copies of some sequence in a genome, independent of their location. The "sequence level" representation here supports this use case. By contrast, a "feature level" representation, where identity of a copy number complement would be based on the identity of member *features*), does not - because we have two sets comprised of entirely different features (based on location being tied to their identity).