{
  "id": "copy-number-variation",
  "title": "Copy-number variation",
  "lang": "en",
  "source_lang": "en",
  "summary": "Copy-number variation is a difference in the number of copies of a DNA segment, contributing to genetic diversity, evolution, inherited disorders, and cancer.",
  "category": "science",
  "subcategory": "biology",
  "type": "concept",
  "aliases": [
    "Copy number variation",
    "CNV"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/copy-number-variation/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/copy-number-variation/",
      "json": "https://aiwiki.page/zh/copy-number-variation.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Copy-number variation is a form of genetic diversity in which a segment of DNA occurs in different numbers of copies among individuals, chromosomes, or cells. It includes deletions, duplications, and more extensive amplifications, ranging from relatively small segments to regions containing many genes. Copy-number variants are a major class of genome structural variation: unlike a change that substitutes one DNA base for another, they alter the amount of particular genetic material.\n\nCopy-number changes can be inherited, arise during the formation of reproductive cells, or develop after fertilization in only some cells of an organism. Many have little detectable effect; others influence gene expression, contribute to adaptation, cause developmental or neurological disorders, or promote cancer. Their consequences depend not simply on segment size but also on which sequences are affected, how copies are arranged, and the biological context.\n\nResearch on copy-number variation connects chromosome biology, molecular mechanisms of mutation, population genetics, and clinical genomics. Its study has also changed the conception of a reference genome: a single linear sequence cannot fully represent the variable numbers and arrangements of DNA segments found within a species."
    },
    {
      "heading": "Definition, scale, and genomic architecture",
      "text": "The term copy-number variation describes a difference in the abundance of a defined DNA sequence relative to another genome or to an established baseline. In a typical diploid human cell, most autosomal regions have two copies, one on each homologous chromosome. A deletion affecting one homolog usually reduces the total to one; a duplication affecting one homolog usually raises it to three. These expectations change for sex chromosomes, normally variable gene families, and organisms or tissues with different chromosome complements.\n\nDefinitions have evolved with measurement technology. Early genome-wide studies commonly used a lower size threshold of approximately 1 kilobase. Many contemporary structural-variant frameworks use 50 base pairs as an operational boundary, so copy-number changes can now include substantially smaller events. Neither threshold represents a fundamental biological discontinuity. “Copy-number polymorphism” generally denotes a relatively common variant, historically often one with a population frequency of at least 1%, whereas “copy-number variant” does not itself imply either commonness or disease.\n\nA copy-number state is not a complete description of chromosome structure. Three total copies could result from a tandem duplication, an additional copy inserted elsewhere, or a more complex rearrangement. Similarly, a total of two copies does not establish that both homologs are intact: one chromosome could carry a deletion and the other a duplication. The arrangement of changes on each homolog—their haplotype context—can therefore matter independently of the total count. Multiallelic loci have more than two copy-number states in the population, sometimes spanning a broad range.\n\nCopy-number variants belong to the broader category of structural variation, which also includes balanced inversions and translocations that need not alter DNA dosage. At the largest scale, aneuploidy changes the number of whole chromosomes, while polyploidy changes the number of complete chromosome sets. These phenomena share dosage consequences with segmental copy-number changes but are usually classified separately. At the smallest scale, short repeat expansions overlap conceptually with copy-number change, although they are generally analyzed using specialized terminology and methods.\n\nGenomic context strongly affects both biological behavior and detectability. Segmental duplications are long, highly similar blocks of sequence present at multiple genomic locations; they are architectural features rather than necessarily variable regions. Some remain stable, while others predispose nearby DNA to recurrent rearrangement. A reference assembly may contain fewer or more copies than another genome, so a “deletion” or “duplication” relative to that reference is not automatically an evolutionary loss or gain. Establishing the ancestral state requires comparative evidence rather than reference comparison alone."
    },
    {
      "heading": "Discovery and population-scale mapping",
      "text": "Large chromosome deletions and duplications were recognized through cytogenetics before the term copy-number variation became widespread. Microscopic chromosome analysis connected altered chromosome structure to human developmental conditions, but its resolution was insufficient to reveal most small or intermediate-sized variants. Molecular investigations subsequently identified disease-associated rearrangements and variable gene families, demonstrating that DNA dosage could differ below the scale visible in conventional chromosome preparations.\n\nThe decisive change came when genome-wide surveys showed that apparently healthy individuals also carried numerous gains and losses. In 2004, independent studies led by Jonathan Sebat and by A. John Iafrate used array-based approaches to identify widespread large-scale variation in human DNA copy number. These findings challenged the assumption that human genomes differed mainly through single-base substitutions and that most sizeable dosage alterations were necessarily exceptional or pathogenic. In 2006, Richard Redon and colleagues extended genome-wide mapping across 270 individuals from populations represented in the International HapMap Project.\n\nEarly maps required careful interpretation. A “copy-number-variable region” was often defined by merging overlapping variants from different people. Its total genomic span therefore did not equal the amount of altered DNA carried by any one individual. Different platforms also sampled different size ranges and genomic environments. Consequently, estimates of how much of the human genome was copy-number variable depended strongly on array design, sample composition, variant definitions, and the treatment of overlapping events.\n\nLarge sequencing projects subsequently integrated copy-number variants with other forms of genomic variation. The 1000 Genomes Project, whose phase 3 analysis appeared in 2015, provided a major population reference for sequence and structural variants. Later resources, including the structural-variation component of the Genome Aggregation Database, improved estimates of population frequency and identified regions in which disruptive variation is unusually depleted. Such depletion can indicate functional constraint, although it must be distinguished from poor detection in repetitive sequence.\n\nPopulation mapping has revealed a heterogeneous landscape rather than a uniform distribution of events. Some loci carry recurrent deletions or duplications with similar boundaries, while others contain many structurally distinct alleles. Common multiallelic regions often contain genes involved in environmental interactions, including immune responses and metabolism. Rare variants are more likely than common variants to include strongly deleterious events, partly because selection limits their frequency.\n\nCatalogs remain dependent on which populations and technologies are represented. A variant absent from a database is not necessarily newly arisen or harmful, and apparent frequency differences may reflect technical ascertainment. Long-read sequencing and the development of the human pangenome are extending these maps into repetitive and structurally complex regions that earlier reference-based analyses represented incompletely."
    },
    {
      "heading": "Molecular mechanisms of formation",
      "text": "Copy-number variants arise through several mechanisms involving chromosome alignment, DNA break repair, or errors during DNA replication. No single pathway explains all events. The genomic sequence surrounding a rearrangement, especially the structure of its junctions, can provide evidence about its origin, but similar final structures may sometimes arise by different routes.\n\nA major mechanism is non-allelic homologous recombination. Normal homologous recombination exchanges information between corresponding DNA sequences. When highly similar repeated blocks at different genomic positions misalign, recombination can instead occur between non-equivalent sites. Directly oriented repeats can generate deletions and reciprocal duplications; the precise products depend on whether the exchange occurs within one DNA molecule or between chromatids or chromosomes. Repeats in opposite orientation more commonly promote inversions. Such processes can occur during meiosis or in mitotically dividing cells.\n\nThis mechanism explains many recurrent genomic disorders. Individuals unrelated by recent ancestry can carry rearrangements of approximately the same interval because the same flanking repeats repeatedly provide substrates for misalignment. Breakpoints may lie within long, nearly identical sequences, making their exact positions difficult to determine. Recurrence therefore does not imply descent from a single ancestral mutation. Conversely, a shared ancestral copy-number variant may persist through many generations without being recreated.\n\nDNA double-strand breaks provide another route. DNA repair by end-joining pathways can reconnect ends after an intervening segment has been lost or can assemble DNA fragments in abnormal combinations. Classical nonhomologous end joining does not require extensive sequence identity. Other repair processes can use short matching sequences, termed microhomology. Junctions containing small insertions or a few shared bases are consistent with such pathways, although microhomology alone is not definitive evidence for one particular mechanism.\n\nReplication-based mechanisms can generate especially complex patterns. A stalled or broken replication fork may resume synthesis using a different template, sometimes switching templates more than once. Models such as fork stalling and template switching and microhomology-mediated break-induced replication account for rearrangements containing duplications, triplications, inversions, and inserted fragments derived from distant positions. Rather than producing a simple loss or gain, one episode can create an intricate series of copy-number transitions.\n\nOther processes operate at particular loci or in particular cellular settings. Slippage during replication can alter repeat number, while transposable elements can introduce new DNA copies or provide homologous sequences that facilitate rearrangement. In cancer, chromosome bridges, breakage–fusion–bridge cycles, and catastrophic chromosome fragmentation can generate extensive dosage changes. Mutation rates consequently vary greatly across the genome: highly rearrangement-prone regions cannot be treated as though they share the same probability of mutation as unique, structurally stable sequence."
    },
    {
      "heading": "Effects on gene function and regulation",
      "text": "The most direct consequence of copy-number variation is altered gene dosage, but a change in DNA copy number does not necessarily produce an equal proportional change in RNA or protein. Regulatory feedback, transcriptional control, RNA turnover, and protein degradation can buffer or amplify the effect. Different tissues may also respond differently because they express different genes and regulatory programs.\n\nA deletion can cause haploinsufficiency when one functioning copy does not provide enough gene product for normal biological activity. Conversely, an additional intact copy of a dosage-sensitive gene can be harmful through increased expression, a property often termed triplosensitivity. Neither property applies to all genes. Many tolerate loss of one copy or modest increases in dosage, whereas others participate in pathways or multiprotein complexes that require tightly controlled relative quantities.\n\nCopy-number variants can disrupt a gene without changing its dosage in the straightforward sense. A breakpoint within a coding sequence may destroy normal function, and duplication of selected exons may alter the reading frame or produce an abnormal transcript. The orientation and position of a duplicated segment determine whether it adds a functional copy, interrupts another gene, or creates a fusion. Consequently, a reported “gain” that overlaps a gene cannot automatically be interpreted as increased production of its normal protein.\n\nChanges outside coding regions can also be consequential. Deletions may remove a promoter or distant enhancer, while duplications may increase the number of regulatory elements. Rearrangements can alter chromatin organization and bring a gene under the influence of regulatory sequences that normally act elsewhere. Disruption of boundaries between regulatory domains can therefore affect genes beyond the nominal copy-number interval. These positional effects help explain why variants of similar size can have very different consequences.\n\nCopy-number change can interact with variation on the other homolog. A deletion may expose a recessive pathogenic allele by removing the corresponding functional copy. Effects can also depend on parent of origin at regions subject to genomic imprinting, where expression differs between maternally and paternally inherited copies. Total copy number alone is then insufficient to establish functional dosage.\n\nAt the organismal level, the relationship between copy number and phenotype may be nonlinear. A small increase may be tolerated until a physiological threshold is crossed, while changes affecting multiple interacting genes can have combined effects. Genetic background, developmental timing, and environment further modify outcomes. Thus, molecular interpretation requires attention to the exact affected sequences and their regulation, rather than assuming that larger variants are always more consequential or that duplications are simply the inverse of deletions."
    },
    {
      "heading": "Evolutionary roles and variation across species",
      "text": "Copy-number variation supplies material for evolution by changing both gene dosage and the number of sequences available for subsequent modification. A duplicated gene can retain its original function while another copy accumulates changes. Over longer periods, duplicated copies may divide ancestral functions between them, acquire new functions, or lose function. Most new duplicates do not necessarily persist: their fate depends on selection, chance, genomic context, and the costs of altered dosage.\n\nNatural selection acts differently on different classes of copy-number change. Strongly harmful variants tend to remain rare or disappear, whereas nearly neutral variants can change frequency through genetic drift. Some variants confer an advantage under particular environmental conditions. Population frequency alone cannot establish which process predominates, because demographic history and recurrent mutation can generate patterns resembling selection.\n\nThe human amylase region illustrates both the biological interest and the analytical complexity of variable copy number. The salivary amylase gene, AMY1, occurs in different copy numbers among individuals. A widely cited 2007 study reported higher average AMY1 copy number in populations traditionally consuming starch-rich diets, providing evidence consistent with dietary adaptation. Subsequent work has shown that the region contains complex structural haplotypes and that copy-number measurement can be technically difficult. Its evolutionary history therefore requires joint analysis of sequence, structure, population history, and expression rather than a simple equation between diet and copy count.\n\nIn other organisms, experimentally supported examples connect amplification to environmental resistance. Increased copy number of the EPSPS gene can contribute to glyphosate resistance in Palmer amaranth, a major agricultural weed. Amplification of the pfmdr1 locus in the malaria parasite Plasmodium falciparum is associated with altered susceptibility to several antimalarial drugs, including mefloquine. In laboratory yeast, nutrient-limited growth can favor amplification of genes encoding relevant transport proteins. These cases show how dosage changes can provide a rapid route to adaptation without first requiring a new protein sequence.\n\nCopy-number dynamics are also important in bacteria. Tandem amplifications can arise and disappear relatively quickly, sometimes producing unstable increases in resistance or metabolic capacity. The fitness benefit under selection may be offset by the cost of carrying and expressing additional DNA when the environmental pressure is removed. Such changes can also increase the number of gene copies in which a subsequent advantageous sequence mutation might occur.\n\nAcross species, comparisons must distinguish polymorphic variation within populations from duplications that have become fixed. A lineage-specific gene-family expansion may have originated through copy-number mutations but no longer vary within the species. Conversely, similar total counts in two species can conceal different histories of duplication and loss. Integrating structural data with phylogenetic relationships is necessary to reconstruct those histories."
    },
    {
      "heading": "Constitutional variation and human genetic disorders",
      "text": "Constitutional copy-number variants are present from conception or arise sufficiently early in development to occur across much of the body. They may be inherited or arise de novo. Their clinical effects range from none detectable to specific Mendelian disorders or broader developmental syndromes. A finding can be medically important even if it is smaller than many common benign variants, because the identity and dosage sensitivity of the affected genes are more informative than size alone.\n\nCharcot–Marie–Tooth disease type 1A provides a well-established example of a duplication disorder. Most cases involve a recurrent duplication of approximately 1.4 megabases at chromosome 17p12 that includes PMP22. Increased PMP22 dosage disrupts peripheral nerve myelin biology. The reciprocal deletion, involving the same region and reduced PMP22 dosage, commonly causes hereditary neuropathy with liability to pressure palsies. Recognition of the CMT1A duplication in 1991 was a landmark in understanding disease caused by gene dosage.\n\nOther disorders demonstrate the effects of deleting intervals containing several genes. The recurrent 22q11.2 deletion typically spans approximately 2.5–3 megabases in its common form and is associated with variable combinations of cardiac, palatal, immune, developmental, and other features. Smaller nested deletions also occur. Williams–Beuren syndrome commonly involves deletion of roughly 1.5–1.8 megabases at 7q11.23, including ELN; loss of ELN contributes to the characteristic vascular abnormalities. These are contiguous-gene disorders, in which different components of the phenotype can reflect loss of different genes within one interval.\n\nCopy-number changes also contribute to conditions traditionally discussed as single-gene disorders. Deletions and duplications of one or more DMD exons are major causes of dystrophin-related muscular disease. Deletion of alpha-globin genes is a common molecular basis of alpha-thalassemia. In these examples, exon structure, reading frame, and the number of remaining functional copies help determine the biological outcome. A whole-region copy count may therefore be less informative than detailed gene-level characterization.\n\nSome recurrent CNVs, including changes at 16p11.2 and other genomic intervals, are associated with increased likelihood of developmental, neurobehavioral, or psychiatric phenotypes without producing a uniform syndrome in every carrier. Two key concepts are penetrance, the proportion of carriers expressing a specified phenotype, and variable expressivity, the range of manifestations among affected carriers. An apparently unaffected transmitting parent does not by itself establish that a variant is harmless, but inheritance is relevant evidence.\n\nClinical association is not equivalent to a deterministic prediction for an individual. Ascertainment through specialist clinics can overrepresent severe presentations, while population studies may miss subtle or age-dependent effects. Interpretation therefore distinguishes a demonstrably dosage-sensitive disease mechanism from a statistical association with variable outcomes, and separates the classification of a variant from its contribution to a particular person's presentation."
    },
    {
      "heading": "Somatic copy-number change and cancer",
      "text": "A copy-number change arising after fertilization can produce mosaicism, in which genetically distinct cell populations coexist within one individual. The distribution of the variant depends on when it occurred and which cell lineage acquired it. An early developmental event may affect several tissues, whereas a later event can be restricted to a small tissue compartment. A sample from blood therefore need not represent the copy-number state of every organ.\n\nSomatic copy-number alterations are especially prominent in cancer. Tumor cells may acquire chromosome-arm gains, focal amplifications, deletions, and whole-genome duplications. Some changes promote growth by increasing oncogene dosage or reducing the activity of tumor suppressor genes. Others accompany chromosome instability without providing an obvious selective advantage. Distinguishing driver alterations from passengers requires evidence beyond their mere presence in a tumor.\n\nWell-established examples include ERBB2 amplification in subsets of breast and gastric cancers, MYCN amplification in a subset of neuroblastomas, and EGFR amplification in some glioblastomas. Deletions involving tumor suppressors such as CDKN2A can remove restraints on cell-cycle progression. Copy-number status can contribute to tumor classification or biomarker assessment, but its significance is specific to the gene, tumor type, molecular context, and method of measurement.\n\nAmplification can involve unusual chromosome structures. Repeated breakage and fusion of chromosome ends can create progressively amplified segments. Alternatively, oncogenes may be carried on extrachromosomal DNA, often as circular molecules lacking conventional centromeres. Their segregation can be uneven, producing substantial cell-to-cell variation in copy number. Catastrophic rearrangement processes such as chromothripsis can shatter and reassemble chromosomes, generating patterns of segmental loss and retention and, in some cases, contributing to amplification.\n\nTumor measurements are complicated by mixtures of malignant and nonmalignant cells, differing tumor ploidy, and subclonal diversity. Four copies of a region may constitute a relative gain in one tumor but an ordinary baseline after whole-genome duplication in another. Likewise, an observed intermediate signal can reflect either a modest alteration in most cells or a larger alteration in a minority of cells. Copy-number profiles therefore require estimates of tumor purity, baseline ploidy, and the fraction of cells carrying each event.\n\nNot all somatic copy-number change indicates cancer. Mosaic chromosome alterations can occur in blood cell populations and become more frequent with age, sometimes reflecting expansion of a clone. Normal immune-cell development also includes programmed DNA rearrangements, such as V(D)J recombination, that can remove intervening DNA. These biological processes must be distinguished from constitutional variation, technical artifacts, and disease-associated acquired alterations."
    },
    {
      "heading": "Detection and measurement technologies",
      "text": "No single method detects all copy-number variants equally well. The appropriate analytical framework depends on variant size, sequence repetitiveness, sample composition, and whether the objective is genome-wide discovery or measurement of a particular locus. Crucially, measuring DNA abundance is not always equivalent to reconstructing the underlying chromosome structure.\n\nChromosomal microarrays transformed genome-wide detection. Comparative genomic hybridization compares labeled test and reference DNA hybridized to genomic probes. Relative signal intensity indicates gains or losses across the interrogated regions. Single-nucleotide polymorphism arrays add allele-specific measurements, which help distinguish copy-number states and identify long regions of homozygosity or allelic imbalance. Arrays generally cannot reveal balanced rearrangements and may not determine where an extra copy has been inserted.\n\nFor an idealized diploid sample compared with a two-copy reference, a simple expected intensity measure is\n\n\\[R=\\log_2\\left(\\frac{C}{2}\\right),\\]\n\nwhere \\(C\\) is total copy number. A one-copy deletion gives \\(R=-1\\), while three copies give \\(R\\approx0.585\\). Real measurements deviate because of background signal, probe behavior, normalization, and cell mixtures. A homozygous deletion cannot be represented experimentally by the formula's mathematical limit because assays retain noise and nonspecific signal.\n\nDNA sequencing provides several complementary signals. Read depth estimates abundance from the number of reads mapping to a region. Discordant paired-end reads suggest unexpected distances or orientations between fragments. Split reads span rearrangement junctions, and local assembly can reconstruct the altered sequence. Combining these signals improves accuracy, but short reads often map ambiguously in segmental duplications and other repeats. Exome sequencing can detect many exon-level changes, although uneven target capture and sparse coverage outside coding regions limit interpretation.\n\nLong-read sequencing can span repeats and connect duplicated copies to unique flanking sequence, improving structural reconstruction and haplotype assignment. Genome assembly can identify sequences missing from the reference and distinguish architectures that produce similar depth signals. Optical genome mapping examines long labeled DNA molecules and is useful for many large rearrangements, although it does not provide base-level sequence and has its own size and labeling constraints.\n\nTargeted methods remain important. Quantitative polymerase chain reaction, digital PCR, and multiplex ligation-dependent probe amplification measure selected loci or exons. Fluorescence in situ hybridization can help establish the chromosomal location and cellular distribution of large changes. Independent methods are particularly informative when a result lies near an assay's detection limit or when structural interpretation matters. Performance must be characterized in terms of sensitivity, specificity, breakpoint precision, and measurement uncertainty, rather than by a nominal resolution alone."
    },
    {
      "heading": "Interpretation, computational challenges, and emerging approaches",
      "text": "Computational analysis converts noisy molecular measurements into discrete copy-number calls. A typical workflow normalizes signal, segments the genome into regions with similar measurements, and assigns likely copy-number states. Hidden Markov models, change-point methods, likelihood approaches, and machine-learning systems are used for different technologies. All must account for factors such as GC content, mapping ambiguity, batch effects, and sample quality. An apparent gain can otherwise reflect technical coverage bias rather than additional DNA.\n\nBenchmarking is difficult because truth sets are least complete in the regions where detection is hardest. Different tools may describe the same complex rearrangement as one event or several, and their breakpoint coordinates can differ even when they agree on the affected genes. Performance estimates therefore depend on matching rules, size thresholds, genomic regions evaluated, and whether the benchmark includes multiallelic or mosaic variants. High accuracy on simple deletions does not establish equivalent accuracy on complex duplications.\n\nClinical classification combines several kinds of evidence: overlap with established dosage-sensitive genes or regions, population frequency, functional consequences, inheritance, segregation, and similarity to previously characterized cases. Joint recommendations published by the American College of Medical Genetics and Genomics and the Clinical Genome Resource in 2020 formalized an evidence-based framework for constitutional CNVs. The framework uses five categories—pathogenic, likely pathogenic, uncertain significance, likely benign, and benign—and separates variant classification from interpretation in the individual clinical context.\n\nLarge-scale association studies face related challenges. Copy-number alleles can be multiallelic, recurrent, poorly tagged by nearby single-nucleotide variants, or correlated with several other structural changes. A genome-wide association study may therefore implicate a region without identifying the causal dosage change. Reliable analysis requires appropriate ancestry matching, technical harmonization, and, where possible, direct structural genotyping. Associations between copy number and expression or phenotype also require attention to confounding and to the possibility that a linked sequence variant is responsible.\n\nReference design is becoming an increasingly important part of the solution. Linear-reference methods can misclassify copies that are absent, collapsed, or differently arranged in the reference assembly. Pangenome graphs and haplotype-resolved assemblies can represent alternative structures explicitly, making it possible to distinguish copy-number states that previously appeared equivalent. Their use also creates new needs for stable identifiers, interoperable coordinates, and consistent reporting of structurally complex alleles.\n\nSingle-cell assays and spatially resolved methods are extending analysis from an average tissue measurement to the distribution of copy-number states among cells. Sparse coverage and amplification artifacts remain substantial limitations, especially for small events. Integrating such data with RNA expression, chromatin measurements, and experimentally engineered variants offers a way to test dosage mechanisms directly. The central interpretive problem is increasingly not whether DNA quantity differs, but which structural allele is present, in which cells, and how that allele changes biological function."
    }
  ],
  "links": [
    "genetic-diversity",
    "dna",
    "genome",
    "gene-expression",
    "cancer",
    "chromosome",
    "aneuploidy",
    "cytogenetics",
    "pangenome",
    "dna-replication",
    "homologous-recombination",
    "meiosis",
    "dna-repair",
    "transposable-element",
    "gene",
    "protein",
    "promoter-biology",
    "enhancer-biology",
    "chromatin",
    "allele",
    "genomic-imprinting",
    "phenotype",
    "evolution",
    "natural-selection",
    "genetic-drift",
    "bacteria",
    "phylogenetics",
    "thalassemia",
    "penetrance",
    "mosaicism",
    "vdj-recombination",
    "comparative-genomic-hybridization",
    "dna-sequencing",
    "polymerase-chain-reaction",
    "measurement-uncertainty",
    "genome-wide-association-study"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}