Chromosomal microarray analysis (CMA) is a laboratory method that examines many locations across the genome simultaneously to identify gains and losses of DNA. These changes range from whole-chromosome abnormalities to deletions and duplications too small to be recognized by conventional chromosome microscopy. Some platforms also reveal extended regions in which the two chromosome copies carry identical genetic markers, providing information about inheritance, chromosome number, and certain forms of mosaicism.
CMA connects molecular genetics with cytogenetics. It has become an important diagnostic technique for developmental disorders and congenital anomalies, prenatal investigation, and the characterization of acquired abnormalities in cancer. Its principal strength is genome-wide assessment of DNA dosage without requiring a suspected chromosomal location. Its interpretation nevertheless depends on the assay design, the biological specimen, and evidence about the clinical significance of the detected variation.
Biological basis and detectable variation
The central target of CMA is copy-number variation: a difference in the number of copies of a genomic segment relative to an expected baseline. Most autosomal regions normally occur in two copies in a human diploid cell. A deletion may reduce a region to one copy, whereas a duplication may increase it to three. Whole-chromosome gains or losses produce aneuploidy, while changes involving an entire additional chromosome set constitute polyploidy. The ability to distinguish these conditions depends on whether the assay measures only relative DNA abundance or also examines genetic-marker patterns.
Copy-number changes differ widely in biological effect. Many are common population variants with no established disease association. Others alter a dosage-sensitive gene, interrupt its coding sequence, remove regulatory elements, or affect several genes simultaneously. A deletion can cause disease through haploinsufficiency, in which one functional copy does not provide sufficient gene activity. Some duplications produce disease because additional copies increase gene activity, although the relationship between DNA dosage and gene expression is not uniformly proportional.
Recurrent genomic disorders often arise through nonallelic homologous recombination. Similar DNA sequences located at different positions can misalign during meiosis, producing a deletion in one recombination product and a reciprocal duplication in another. This mechanism explains why unrelated individuals may carry rearrangements with similar boundaries. The recurrent 22q11.2 deletion and the approximately 1.5–1.8-megabase deletion associated with Williams–Beuren syndrome illustrate clinically important abnormalities that can be detected by CMA without a separate locus-specific test.
Other rearrangements have unique boundaries and arise through mechanisms involving DNA breakage, repair, or replication. A copy-number change may encompass part of a gene, a single gene, or many millions of DNA bases. Consequently, size alone is not a reliable indicator of pathogenicity: a small deletion affecting a dosage-sensitive gene can be more consequential than a much larger change in a relatively gene-poor region.
On platforms containing suitable markers, CMA also detects regions of homozygosity, sometimes described as absence of heterozygosity. These regions contain no detectable difference between the two inherited copies at informative markers. Multiple long regions distributed across the genome may reflect shared parental ancestry. A large region restricted to one chromosome may raise the possibility of uniparental disomy, in which both homologous chromosome copies originate from one parent. Such a pattern is evidence for further interpretation, not by itself proof of parental origin or a specific disorder.
Array platforms and signal generation
Two major technological approaches underpin clinical CMA: array comparative genomic hybridization and arrays based on single-nucleotide polymorphisms. Both use large collections of immobilized DNA probes, but they extract different information from the DNA that binds to those probes. Modern clinical platforms may combine copy-number probes with polymorphic markers, reducing the practical distinction between these categories.
In array comparative genomic hybridization, commonly abbreviated array CGH or aCGH, patient and reference DNA are typically labeled with different fluorescent dyes. The samples are mixed and allowed to hybridize to the array. Each probe captures complementary DNA from its corresponding genomic region. After washing and scanning, the relative fluorescence intensities estimate the abundance of that region in the patient compared with the reference.
A simplified expression for the dosage signal is
Under idealized conditions, equal copy number gives . A one-copy deletion against a two-copy reference gives , while a three-copy duplication gives . Actual observed values are affected by normalization, probe response, background signal, and mixtures of cells. These theoretical values therefore describe the underlying relationship rather than fixed diagnostic thresholds.
SNP arrays measure total signal intensity together with the relative contribution of alternative alleles at polymorphic sites. A commonly displayed quantity is B-allele frequency, which estimates the fraction of signal attributable to the allele designated B. In a normal diploid region, the principal genotype bands lie near 0, 0.5, and 1, corresponding to AA, AB, and BB. A three-copy region can produce additional bands near one-third and two-thirds, whereas a deletion removes the ordinary heterozygous band.
Allelic information makes SNP-containing arrays capable of detecting copy-neutral abnormalities that a dosage-only assay cannot reveal. For example, a region can retain two copies while both copies derive from the same ancestral chromosome segment. SNP patterns can also support identification of triploidy, which relative dosage-only array CGH generally cannot detect reliably when the entire genome is proportionally increased.
Platform resolution is not determined simply by the total number of probes. Probe spacing, the number of consecutive abnormal probes required for a call, local sequence complexity, and laboratory reporting thresholds all matter. Coverage is often deliberately denser within clinically important genes or recurrent disease regions. Two arrays containing similar numbers of probes can therefore have substantially different abilities to detect a particular intragenic deletion.
Specimens, laboratory workflow, and quality control
CMA generally begins with extraction of genomic DNA rather than preparation of dividing cells. Constitutional testing commonly uses peripheral blood, but saliva, buccal specimens, cultured fibroblasts, and other tissues may be suitable under validated laboratory procedures. Prenatal specimens include chorionic villi and amniotic fluid. Tumor analysis may use fresh tissue, bone marrow, or formalin-fixed material, although fixation can substantially degrade DNA.
The specimen determines which cell populations are represented. Blood can fail to reveal a chromosomal abnormality confined mainly to another tissue. Conversely, an acquired blood-cell clone may create an abnormal array pattern that is not representative of the person’s constitutional genome. In suspected mosaic conditions, differences between tissues are therefore biologically important rather than merely technical inconsistencies.
After extraction, DNA concentration, purity, and integrity are assessed. Depending on the platform, processing may include restriction digestion, amplification, fragmentation, labeling, and hybridization. The array is washed and scanned, and software converts fluorescence measurements into normalized intensity and genotype data. Computational segmentation then groups adjacent probes whose signals support a shared copy-number state.
Quality assessment includes the variability of adjacent probe signals, consistency of genotype clusters, the proportion of usable markers, and evidence of contamination or sample mixture. Poor DNA quality, uneven amplification, and genomic waves associated with factors such as GC content can generate false segmentation boundaries. A convincing call generally requires a coherent regional pattern, not an isolated abnormal probe.
Prenatal samples present additional concerns. Maternal-cell contamination can obscure fetal abnormalities or generate misleading mixtures. Chorionic villi represent placental tissue, so an abnormal result may reflect confined placental mosaicism rather than the fetal chromosome constitution. Culturing cells can change the proportions of different cell lines through selective growth. These factors influence how the result is interpreted and whether findings from another specimen are concordant.
Orthogonal testing may be used to resolve a finding’s structure, inheritance, or technical credibility. Fluorescence in situ hybridization can show the chromosomal location of selected sequences; targeted quantitative assays can assess dosage; and conventional chromosome analysis can identify a rearrangement’s broader architecture. Confirmation is not automatically required for every clear result from a validated assay. The relevant issue is whether an additional method answers an unresolved question.
A clinical report typically specifies the reference genome assembly, chromosome coordinates, estimated size, copy-number state, classification, and assay limitations. Boundaries usually fall between the last normal and first abnormal informative probes. They are therefore estimates rather than nucleotide-level breakpoint sequences, even when reported as precise genomic coordinates.
Classification and clinical interpretation
Detecting a copy-number change is an analytical task; deciding what it means is a separate interpretive task. Clinical laboratories consider the affected genes, established dosage sensitivity, observations in affected and unaffected populations, inheritance, and the individual’s phenotype. A technically secure finding can remain clinically uncertain when biological evidence is insufficient.
The American College of Medical Genetics and Genomics and the Clinical Genome Resource published joint technical standards for constitutional copy-number variant interpretation in 2020. These standards use an evidence-based scoring framework and five categories: pathogenic, likely pathogenic, variant of uncertain significance, likely benign, and benign. Separate criteria address deletions and duplications because loss and gain of the same region do not necessarily have equivalent effects.
Gene content is central, but counting genes is not enough. A deletion encompassing a well-established haploinsufficient gene may have strong pathogenic evidence. A duplication involving the same gene requires evidence that increased dosage is harmful or that the rearrangement disrupts gene function. Partial-gene duplications can be particularly difficult to interpret because their effect depends on orientation, insertion site, and whether they alter the reading frame—features that an array may not establish.
Population frequency provides another major line of evidence. A recurrent copy-number variant observed frequently in unaffected individuals is generally less likely to cause a highly penetrant, severe disorder. However, databases differ in the ancestry, clinical assessment, and technical resolution of the people represented. Apparent rarity is consequently not equivalent to pathogenicity, and absence from a database is not sufficient evidence of disease causation.
Inheritance is informative but not decisive. A newly arisen, or de novo, variant may strengthen a pathogenic interpretation when its gene content and phenotype are compatible. An inherited variant may still contribute to disease because of penetrance below 100%, variable expression, or interactions with other genetic factors. An apparently unaffected parent may also have subtle features that were not systematically assessed. Conversely, a de novo finding can be benign.
Interpretation distinguishes the classification of a variant from its relevance to a particular case. A pathogenic deletion associated with one condition may not explain every clinical feature in an individual. Likewise, a variant of uncertain significance is not an established diagnosis. Uncertainty can change as additional cases, segregation data, functional studies, and population observations become available. Reclassification reflects changes in the evidence rather than a change in the underlying DNA result.
Postnatal constitutional diagnosis
CMA became especially important in the investigation of unexplained developmental delay, intellectual disability, congenital anomalies, and some presentations of autism spectrum disorder. These conditions are genetically heterogeneous: similar clinical findings can arise from changes in many different genomic regions. A genome-wide dosage assay can therefore identify abnormalities that would be missed by choosing a single chromosomal region for testing.
A widely cited 2010 consensus statement led by David T. Miller and colleagues reviewed 33 studies involving 21,698 patients. It reported a diagnostic yield of approximately 15–20% for CMA in relevant developmental and congenital-anomaly populations, compared with approximately 3% for conventional chromosome analysis when recognizable chromosome syndromes were excluded. These figures describe selected study populations and contemporary interpretation, not a fixed yield for every indication or laboratory.
Examples of clinically significant findings include deletions or duplications at 16p11.2, 1q21.1, and 22q11.2, as well as larger, nonrecurrent imbalances. Different changes within the same cytogenetic region can have different consequences, making exact coordinates and gene content essential. Even well-characterized recurrent variants may be associated with a broad range of developmental, behavioral, and physical findings rather than one uniform clinical presentation.
CMA can also reveal that an apparently isolated chromosomal change is part of a more complex imbalance. For instance, a terminal deletion on one chromosome accompanied by a terminal duplication on another may suggest an unbalanced translocation. The array identifies the dosage changes, but chromosome analysis or targeted localization may be needed to determine their arrangement and whether a parent carries a balanced rearrangement.
Regions of homozygosity provide a different route to diagnosis. They can identify genomic intervals in which a recessive disease-causing variant is more plausible, particularly when several long regions suggest parental relatedness. The array does not usually identify the causal sequence variant itself. Subsequent DNA sequencing may establish whether a relevant gene within the region contains a pathogenic variant on both copies.
A negative CMA result does not exclude a genetic disorder. Many developmental conditions result from single-nucleotide variants, small insertions or deletions, repeat expansions, or altered DNA methylation, none of which is comprehensively assessed by standard CMA. In 2021, an ACMG evidence-based guideline supported exome or genome sequencing as a first- or second-tier test for pediatric patients with congenital anomalies, developmental delay, or intellectual disability. This broadened the diagnostic framework rather than making dosage analysis biologically irrelevant: sequencing and arrays identify overlapping but nonidentical classes of variation.
Prenatal testing and pregnancy-loss investigation
Prenatal CMA examines DNA from an invasive diagnostic specimen to identify fetal or placental genomic imbalances. It differs fundamentally from screening based on cell-free DNA in maternal plasma: the latter estimates risk for selected abnormalities, whereas CMA directly analyzes DNA obtained from the sampled pregnancy tissue. The array result nevertheless remains subject to specimen identity, contamination, and the possibility that placental and fetal cell populations differ.
Its additional diagnostic contribution is particularly evident when fetal structural abnormalities are present. In a large multicenter study published by Ronald Wapner and colleagues in 2012, CMA identified clinically relevant deletions or duplications in approximately 6.0% of fetuses with a structural anomaly and a normal karyotype. Among pregnancies tested for indications such as advanced maternal age or a positive screening result, the corresponding additional yield was approximately 1.7%. These values depend on the study’s case selection, platform, and classification criteria.
Prenatal interpretation is complicated by incomplete phenotyping. Ultrasound can identify many structural abnormalities but cannot establish later intellectual development, behavioral characteristics, or the full spectrum of age-dependent features. A copy-number variant with variable expression may therefore be more difficult to relate to prognosis before birth than after a detailed postnatal assessment.
Parental testing can help distinguish inherited from de novo changes and clarify recurrence mechanisms, but it does not eliminate all uncertainty. An inherited variant may have different effects in parent and fetus. A de novo variant may remain uncertain if the affected region lacks established clinical evidence. Laboratories also differ in their policies for reporting uncertain findings, susceptibility variants, and findings unrelated to the original indication.
CMA is also used in the investigation of pregnancy loss and stillbirth. Because it does not necessarily require viable dividing cells, it can produce results from tissue in which conventional cell culture fails. It can detect many chromosome-number abnormalities and subchromosomal imbalances, although performance depends on tissue preservation and the assay’s design. Maternal tissue mistakenly sampled as pregnancy tissue can yield a misleadingly normal maternal result.
SNP-based analysis is particularly relevant in this setting because allelic patterns can reveal triploidy and help identify contamination or unusual inheritance. However, not every array platform detects every type of polyploidy or parental-origin abnormality. Some suspected imprinting disorders require targeted methylation testing, parental-marker studies, or other assays. Prenatal CMA therefore supplies a detailed assessment of selected genomic features rather than a comprehensive prediction of fetal health.
Cancer genomes and acquired chromosomal abnormalities
In cancer, CMA is used to characterize acquired genomic changes in tumor cells rather than primarily to identify inherited abnormalities. Malignant genomes frequently contain chromosome gains, losses, focal amplifications, and regions of allelic imbalance. These changes can contribute to tumor development by increasing oncogene dosage, deleting tumor-suppressor genes, or exposing a recessive pathogenic allele.
SNP-containing arrays can detect copy-neutral loss of heterozygosity, a state in which a region retains its overall copy number but loses the normal mixture of parental alleles. One mechanism is mitotic recombination, which can produce a segment containing two copies of the same parental sequence. In a tumor, this can make an existing pathogenic variant effectively homozygous without producing a detectable deletion.
Acquired copy-neutral loss of heterozygosity on chromosome arm 9p is a well-established example in myeloproliferative neoplasms, where it can involve the region containing JAK2. Other recurrent regions occur in myeloid and lymphoid malignancies. The array demonstrates the allelic pattern, while sequence analysis is needed to identify the particular gene variant whose representation has changed.
Copy-number profiles can also reveal deletions involving genes such as CDKN2A or focal amplifications affecting oncogenic regions. The clinical importance of a finding depends on tumor type and the associated evidence. A genomic gain is not automatically equivalent to increased production of the encoded protein, and an array alone does not establish the activity of a signaling pathway.
Tumor purity and clonal composition strongly influence the data. Normal cells dilute abnormal signals, while different tumor subclones may carry distinct copy-number states. A tumor with whole-genome doubling can have a baseline unlike that of a normal diploid specimen. Analysis must therefore consider ploidy, normal-cell admixture, and the possibility that a relative loss still leaves more than two absolute copies.
Without a matched constitutional specimen, some findings cannot be confidently assigned as inherited or acquired. A deletion detected in tumor tissue could be present in all cells or confined to the malignancy. Formalin-related DNA damage and fragmentation add further analytical challenges, although specialized platforms can accommodate many archived specimens.
CMA does not reveal most balanced gene fusions, a major limitation in hematologic malignancies and other tumors. It is consequently used alongside chromosome analysis, targeted FISH, and sequencing-based assays. These methods address distinct questions: genome-wide dosage, chromosomal architecture, specific rearrangements, and nucleotide-level alterations.
Analytical limits and relationships to other methods
CMA measures the abundance and, on suitable platforms, allelic composition of represented genomic sequences. It does not directly reconstruct the physical arrangement of every chromosome. A duplication signal may indicate an extra copy beside the original segment, insertion elsewhere on the same chromosome, or insertion on another chromosome. These configurations can have different biological and reproductive implications despite similar array profiles.
Balanced reciprocal translocations and inversions usually escape detection because they change chromosome structure without changing the amount of DNA. Some apparently balanced rearrangements do contain small deletions or duplications that CMA can reveal, but a normal array cannot establish structural normality. Conventional karyotyping remains informative for large-scale chromosome architecture, while FISH can localize selected sequences.
Detection of mosaicism depends on the fraction of abnormal cells, the size and type of imbalance, marker coverage, and signal quality. There is no single universal percentage below which all arrays fail. In an idealized mixture containing a fraction of cells with a one-copy deletion and a fraction of normal diploid cells, the mean copy number is
The corresponding dosage ratio relative to normal is , so the signal approaches normal as the abnormal fraction decreases. Allelic patterns can provide additional evidence, but tissue sampling remains a fundamental limit.
Standard CMA is also poorly suited to repetitive regions, many centromeric and heterochromatic sequences, and changes smaller than the platform’s effective resolution. A probe located within a gene does not imply complete coverage of all its exons. Laboratories may intentionally exclude highly repetitive sequences because cross-hybridization makes their signals difficult to interpret.
Most single-nucleotide variants and small insertions or deletions require sequencing-based methods. Repeat-expansion disorders and methylation abnormalities often require specialized assays. Even a very dense SNP array examines selected polymorphic positions rather than determining the complete DNA sequence.
Exome and genome sequencing increasingly incorporate copy-number analysis. Genome sequencing can additionally identify many breakpoint-spanning reads and balanced structural variants, while exome-based dosage analysis is concentrated on captured coding regions and can be uneven. Arrays retain advantages in established workflows, predictable genome-wide dosage coverage, and mature allelic analysis. Relative performance depends on the sequencing protocol, computational pipeline, specimen quality, and validation rather than on a simple hierarchy of technologies.
Historical development and standardization
CMA emerged from the convergence of chromosome analysis, comparative genomic hybridization, and DNA microarray technology. Conventional cytogenetics established the importance of visible chromosome abnormalities, but its resolution was constrained by chromosome banding and microscopic appearance. Targeted molecular methods later demonstrated that clinically important deletions could exist below that level of resolution.
In 1992, Anne Kallioniemi and colleagues described comparative genomic hybridization using differentially labeled tumor and reference DNA hybridized to normal metaphase chromosomes. The technique allowed genome-wide identification of relative gains and losses without requiring metaphase chromosomes from the tumor itself. Its spatial resolution was limited by the chromosome substrate and image analysis.
Daniel Solinas-Toldo and colleagues reported matrix-based comparative genomic hybridization in 1997. In 1998, Daniel Pinkel and colleagues described high-resolution analysis of DNA copy-number variation using comparative genomic hybridization to microarrays. Replacing metaphase chromosomes with defined DNA targets allowed probe placement to determine genomic coverage and supported substantially finer localization of imbalances.
Early arrays frequently used large-insert cloned DNA, including bacterial artificial chromosomes. Subsequent platforms adopted synthetic oligonucleotides, enabling much higher probe densities and more flexible designs. In parallel, SNP genotyping arrays developed for genetic mapping and association studies became useful for measuring both DNA dosage and allelic imbalance.
Two independent studies published in 2004, led respectively by Jonathan Sebat and A. John Iafrate, helped establish that substantial copy-number variation is widespread among people without recognized chromosomal disease. This finding altered interpretation fundamentally: a departure from the reference genome could not automatically be treated as pathogenic. Population variation became an essential comparator for clinical testing.
Clinical adoption accelerated during the 2000s as arrays demonstrated diagnostic value in individuals whose conventional chromosome studies were normal. The 2010 consensus statement established an influential framework for CMA as a first-tier cytogenetic diagnostic test in selected developmental presentations. Prenatal studies subsequently quantified its additional yield over karyotyping, while cancer applications developed around the combined analysis of dosage and allelic imbalance.
Standardization has involved more than hardware. Reference genome assemblies, variant nomenclature, laboratory validation, data-sharing practices, and dosage-sensitivity curation all affect whether findings can be compared across institutions. Resources such as ClinGen, DECIPHER, and population structural-variation databases support interpretation, but their entries require evaluation in context. Changes in genomic coordinates between reference assemblies, differences in probe coverage, and evolving clinical classifications remain important when an older result is reassessed.