• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of ajhgLink to Publisher's site
Am J Hum Genet. Jan 2007; 80(1): 91–104.
Published online Dec 5, 2006.
PMCID: PMC1785303

A Comprehensive Analysis of Common Copy-Number Variations in the Human Genome


Segmental copy-number variations (CNVs) in the human genome are associated with developmental disorders and susceptibility to diseases. More importantly, CNVs may represent a major genetic component of our phenotypic diversity. In this study, using a whole-genome array comparative genomic hybridization assay, we identified 3,654 autosomal segmental CNVs, 800 of which appeared at a frequency of at least 3%. Of these frequent CNVs, 77% are novel. In the 95 individuals analyzed, the two most diverse genomes differed by at least 9 Mb in size or varied by at least 266 loci in content. Approximately 68% of the 800 polymorphic regions overlap with genes, which may reflect human diversity in senses (smell, hearing, taste, and sight), rhesus phenotype, metabolism, and disease susceptibility. Intriguingly, 14 polymorphic regions harbor 21 of the known human microRNAs, raising the possibility of the contribution of microRNAs to phenotypic diversity in humans. This in-depth survey of CNVs across the human genome provides a valuable baseline for studies involving human genetics.

Genetic variation in the human genome exists in different forms. Recent studies have shown that variations exist in the human genome at various levels: the single base pair,1 the kilobase pair,24 and tens to thousands of kilobase pairs.58 Extensive studies, including the recently published haplotype map from HapMap,1 have identified millions of SNPs in the human genome. Three recent studies that used the SNP data each identified several hundred sites of deletion in the human population; however, gains could not be deduced from this data set.24 By use of a fosmid paired-end sequence analysis, a comprehensive comparison between two genomes quantified 241 sites of insertion or deletion.8 By use of array comparative genomic hybridization (array CGH) techniques, large-scale copy-number variations (CNVs) were demonstrated in a fraction of the human genome.5,6 Each of these studies added to our knowledge about CNVs in the human population, but with little overlap in findings.9 Thus, many characteristics of CNVs in the human population remain unknown, such as the total number, genomic positions, gene content, frequency spectrum, and patterns of linkage disequilibrium with one another. Understanding CNVs is critical for the proper study of disease-associated changes because segmental CNVs have been demonstrated in developmental disorders and susceptibility to disease.10 Therefore, analysis of CNVs at the whole-genome level is required to create a baseline of human genomic variation. In this study, using a whole-genome tiling-path BAC array CGH approach,11 we measured large scale (>40 kb) segmental gains and losses in >100 individuals to expand our knowledge about CNVs and to estimate the extent of this form of variation in the human population.

Material and Methods

DNA Samples

Samples were collected and were rendered anonymous. These samples included 16 from healthy blood donors, 51 from a British Columbia Cancer Agency (BCCA) screening program, and 26 B-lymphoblast DNA samples encompassing 16 distinct ethnic groups from the Human Variation Collection and 14 CEPH pedigree samples from the Coriell Cell Repository (National Institute of General Medical Sciences, Camden, NJ). The DNA samples from cell lines were included to represent diverse ethnic populations. The 51 samples from the BCCA screening program included 19 from a breast cancer screening program and 32 from a colon cancer screening program. These were constitutional DNA samples obtained from blood that did not contain any neoplastic cells, and none showed CNV association with BRCA1 (MIM 113705), BRCA2 (MIM 600185), APC (MIM 175100), MSH2 (MIM 609309), or MSH6 (MIM 600678). Only 2 of the 32 samples from the colon cancer screening program showed CNV association with MLH1 (MIM 120436). In addition, no CNVs were associated with a specific sample type or source, which suggests no obvious selection bias. In total, 105 DNA samples (from 44 males and 61 females) were included in this study (table 1), 95 of which were used for CNV discovery. DNA from the four grandparents of the CEPH pedigree were included in the CNV discovery sample set, whereas DNA from 10 other members of the family were included only for clustering and inheritance analysis. A donor sample was used as the male reference, and a single female sample was used only in control experiments. Genomic DNA from donors was extracted from whole blood by use of the QIAamp DNA Blood Maxi Kit (QIAGEN) and was quantified by spectrophotometry (ND-1000 [NanoDrop]).

Table 1.
Samples Used in This Study

BAC Array CGH Analysis

DNA labeling and hybridization was performed as described elsewhere,11 with slight modifications. In brief, 200 ng of sample and reference DNA were differentially labeled with Cyanine 3–dCTP and Cyanine 5–dCTP (Perkin Elmer Life Sciences). The random priming reaction was incubated in the dark at 37°C for 16–18 h. DNA samples were then combined, and unincorporated nucleotides were removed using microcon YM-30 columns (Millipore). Purified samples were mixed with 100 μg of human Cot-1 DNA (Invitrogen) and were precipitated. DNA pellets were resuspended in 45 μl of DIG Easy hybridization solution (Roche) containing 20 mg/ml sheared herring sperm DNA and 10 mg/ml yeast tRNA. Sample mixture was denatured at 85°C for 10 min, and repetitive sequences were blocked at 45°C for 1 h before hybridization. The mixture was then applied onto BAC arrays containing 26,363 clones spotted in duplicate (53,856 elements with controls) on single slides. (These clones were selected from the SMRT clone set, to optimize tiling coverage of the genome; the clone list is available at the SMRT Array Web site.11) Hybridization was performed in the dark at 45°C for ~36 h inside a hybridization chamber, followed by washing three times for 3 min each with agitation in 0.1× saline sodium citrate (SSC) and 0.1% SDS at 45°C. Arrays were then rinsed three times for 3 min each in 0.1× SSC at room temperature and were dried by an air stream before imaging. Slides were scanned using a charge-coupled device–based imaging system (arrayWoRx eAuto [Applied Precision]) and were analyzed with the SoftWoRx Tracker Spot Analysis software (Applied Precision). The log2 ratios of the Cyanine 3 to Cyanine 5 intensities for each spot were assessed. To remove systematic effects from nonbiological sources that introduce bias, the ratios were then normalized using a stepwise normalization technique.12 Custom SeeGH software was used to visualize normalized data as log2 ratio plots (fig. 1).13

Figure  1.
Example of a karyogram from a hybridization experiment in this study. Custom SeeGH software was used to visualize normalized data as log2 ratio plots.13 The figure illustrates an example of a hybridization of a female sample versus the male reference. ...

CNV-Detection Algorithm

For each experiment, 1,398 clones from chromosomes X and Y were removed, and the remaining data were median normalized to remove bias introduced because of any sex-mismatched hybridization. In addition, 573 clones were removed from analysis because of printing anomalies or their shift in log2 ratios, possibly due to homology with the X or Y chromosome, leaving a total of 24,392 reliable clones for analysis (see the tab-delimited ASCII file, which can be imported into a spreadsheet, of data set 1 [online only]). Experimental SDs (SDautosome) were calculated for each experiment on the basis of the log2 ratios of the 24,392 reliable clones minus the clones removed because of low signal-to-noise ratio (SNR) or high SD of replicate clone measures (SDclone). Thresholds for determining CNV clones were set at a multiple of the SDautosome value. For each experiment, clones were annotated as uninformative if they were filtered via SNR or SDclone, as a CNV loss if the log2 ratio was less than the negative threshold, as unchanged if the log2 ratio was between the negative and positive thresholds, and as a CNV gain if the log2 ratio was above the positive threshold.

To determine the optimal values for SNR, SDclone, and the SDautosome multiplier, eight hybridization experiments (four repeat experiments of male reference versus the single female DNA and four experiments between those two DNAs and two additional DNA pools) were used. On the basis of the possible combinations of copy-number status in the four DNA samples used, we determined the expected CNV patterns in the eight hybridization experiments (table 2). The three parameters were recursively varied until the highest proportion of CNV clones match the expected patterns (table 2); this resulted in the filter settings of SDclone>0.15, SNR<3, and a stringent SDautosome multiplier of 3.3×. On the basis of six self-versus-self hybridizations to calibrate array performance, experiments with >10% uninformative data points or with an SDautosome>0.12 were repeated. Normalized log2 ratio profiles were generated for the 105 individuals from hybridization of sample DNA versus a single male reference DNA. Data points that did not meet our SDclone or SNR criteria were annotated as uninformative, whereas those whose average ratio exceeded the 3.3× SDautosome were identified as CNV clones (see the tab-delimited ASCII file of data set 2 [online only]). CNV clones that overlapped in genomic coverage were considered to represent the same CNV loci. A custom track file for uploading the identified CNV clones to the University of California–Santa Cruz (UCSC) Human Genome Browser is available on request. After submission of the custom track file, clones displayed in blue, red, green, and black represented CNVs seen once or twice, three times, four or five times, and six or more times, respectively.

Table 2.
Expected CNV Patterns of Eight Hybridizations between Four DNA Samples[Note]

Determination of False-Positive and False-Negative Rates

To estimate our false-positive and false-negative rates in this study, six repeat experiments (of the single female vs. the male reference) were analyzed per our CNV algorithm (see above). In total, 803 CNV calls were made, with 340 seen only once, 50 twice, 46 three times, 15 four times, 15 five times, and 15 six times. Given that our false-positive results cannot exceed the total number of calls (i.e., 803), our maximum false-positive rate is 0.5487% (803/24,392 measures × 6 experiments). By use of this maximum false-positive rate of 0.5487%, the binomial probability, p, of detecting the same clone twice within six experiments by random chance is p=0.000445. Therefore, we concluded that any clone detected twice or more was a true CNV in these six repeat experiments. In theory, we expected to detect 141 true CNVs (i.e., 50 calls seen twice, 46 seen three times, and 15 each seen four, five, and six times) in each of the six experiments (846 calls). In practice, 463 were detected, yielding an estimated false-negative rate of 45.3%. Although statistically a fraction of the single-occurrence calls (those seen only once) represent true CNVs, we conservatively considered all 340 as false-positive results, resulting in a false-positive rate of 0.2323% (340 calls/24,392 measures × 6 experiments). In short, we tolerated this high false-negative rate of 45.3% to achieve our very low false-positive rate for confidence in CNV discovery.

On the basis of the false-positive and false-negative rates calculated above, in a repeat of the same hybridization experiment, one would expect to see 134 calls (803 calls/6 experiments), of which 57 would be false-positive results (0.2323% × 24,392 measures) and 77 would be true CNVs. On the basis of our false-negative rate, we would have missed 64 true CNVs (of 141 true CNVs). Therefore, of a total of 141 true CNVs, the probability of obtaining the same true CNVs in a repeat hybridization should be 54.7% (77 of 141), and the probability of seeing those same CNVs in a second repeat hybridization would be 54.7% × 54.7% (42 of the 141 true CNVs). This represents 84 calls (2 × 42 CNVs) of the 268 expected total calls (134 × 2) (a 31.3% overlap). To verify our calculated rates, three repeat hybridization experiments were performed using the same samples. The observed overlaps of CNV calls between the three possible comparisons were 31.3%, 28.6%, and 31.2%, which is in complete agreement with the expected value. The above calculations are summarized in figure A1 (online only). Additionally, 20 samples (F1, F2, F3, S1, S3, S4, S7, S8, S10, S11, S12, S14, S16, S17, S33, S38, S39, S40, S41, and S44) from the discovery set were each repeated once with a fluorochrome reversal. The overlapping calls between repeats ranged from 21% to 46%, with an average of 30%, again consistent with the expected value from our false-positive and false-negative rates.

Furthermore, we employed an additional platform to verify our CNV calls. We recognize that oligonucleotide arrays are generally not designed for measuring CNVs in certain loci, since many segmental duplications and repeat sequences are excluded from array design, and thus we constructed a custom oligonucleotide array (NimbleGen Systems) covering our 3,654 CNV loci with 389,027 elements (~2 kb spacing between elements). Five samples (S70, S71, S72, S73, and S80) were assayed using this custom platform. Each of these DNA samples were hybridized against the same single male reference DNA used for BAC array analysis onto the oligonucleotide array. As described elsewhere,14 to identify gains or losses from the oligonucleotide array, thresholds of 2 SDs of the mean log2 ratio for all elements in the hybridization were used. On the basis of the detection sensitivity of BAC array CGH,15 a moving window size of 19 elements (for a total of ~40 kb, with ~2 kb spacing between elements) was applied. In each window, the number of elements reporting a loss (beyond the threshold) was subtracted from the number of elements reporting a gain. The difference was then divided by 19—the total number of elements in the window. Gains or losses were scored for results at >0.1 or <−0.1, respectively. Calls from the oligonucleotide array were then directly compared with CNVs detected by BAC array analysis. To confirm a BAC CNV gain (or loss), at least 10 gains (or losses) were required from the oligonucleotide probe calls covering the same BAC.

CNV Association

To obtain the genomic loci of our identified copy-number–altered clones, we used UCSC May 2004 mapping annotations from BACPAC Resources. For comparison, locations of previously identified CNVs obtained from the Database of Genomic Variants and from various publications were also anchored to the UCSC May 2004 assembly (from UCSC Genome Bioinformatics).24 These were then converted to elements (i.e., clones) within our clone set by comparison of chromosome number, base-pair start position, and base-pair end position.

RefSeq gene information was downloaded from the UCSC May 2004 assembly and was viewed in relation to our CNVs. A gene with any overlap across a CNV boundary was considered to be associated with the CNV. Genes overlapping our CNVs were then used to match genes downloaded from the Online Mendelian Inheritance in Man (OMIM) Morbid Map. The locations of human microRNAs were downloaded from the Sanger miRBase database, were converted to the UCSC May 2004 mapping annotations, and were viewed in relation to our CNVs as described above.16

Duplication Analysis

BAC clones and segmental duplication data were mapped to the UCSC May 2004 assembly. CNV loci were assessed for duplication content on the basis of whole-genome assembly comparison (WGAC) and whole-genome shotgun sequence detection (WSSD) analyses of human and chimpanzee genome assemblies.1720 We required >10 kb of duplicated sequence to consider a BAC as duplicated. Lineage-specific duplications were distinguished on the basis of human and chimpanzee-only comparisons,19 available at the Segmental Duplication Database.

Clustering Analysis

A total of 105 individuals were clustered on the basis of our CNV clones, including 14 members of a CEPH pedigree: 4 grandparents (already part of our 95-sample CNV discovery set), 2 parents, and 8 offspring. All clones with copy-number gains and losses were annotated as +1 and −1, respectively. Uninformative measures were left blank, whereas the remaining cells were annotated as 0. Hierarchical clustering of the samples with single linkage was performed using Cluster and was visualized using Treeview21 (Eisen Lab: Software Web site).

Sample Diversity

The diversity between every possible pair of individuals was calculated by enumerating the number of CNVs (observed at least three times among the 95 samples) with differing status. The pair with the largest value was taken to be the most diverse.

Variation in genome size was determined by first enumerating the net gain or net loss of clones (observed at least three times among the 95 samples) within each individual compared with our reference. The maximum variation was calculated by adding the lowest net loss and the highest net gain. To convert this difference in net clones to genomic size, the number of clones was multiplied by the minimum detection sensitivity of BAC array technology, previously shown to be 40 kb for the average-sized BAC clone.15

Quantitative PCR

The iQ SYBR Green Supermix system (Bio-Rad) was used for quantitative PCR (qPCR). Primers were designed using Primer3,22 and the primers tested are summarized in the tab-delimited ASCII file of data set 3 (online only). In brief, 10 ng genomic DNA was used in a 25-μl reaction with a test or reference primer pair at 600 nM. Reactions were performed in triplicate and were repeated on different days by use of a Bio-Rad iCycler Optical Module (at 95°C for 10 min, then 40 cycles at 95°C for 15 s and 60°C for 1 min, followed by final extension 55°C for 1 min and a melting-curve analysis). Standard curves for each primer pair were generated using a 10-fold dilution series ranging from 0.1 ng to 100 ng. Data analysis was performed as described by Weksberg et al.23


Identification of CNVs

By application of a whole-genome tiling-path BAC array CGH technique, pairwise comparison of DNA samples from 95 unrelated individuals against a single reference DNA sample identified a total of 14,711 CNV BAC clones, averaging 155 per individual (array CGH data for all hybridization experiments have been made publicly available at the Gene Expression Omnibus [series accession number GSE5442]). This resulted in 5,132 unique clones that span 3,654 loci throughout the mapped autosomes (fig. 2 and the tab-delimited ASCII file of data set 2 [online only]). To determine a confidence level for our CNVs, we first calculated the probability of an event occurring repeatedly within our sample set. On the basis of our false-positive rate of 0.23%, calculated from repeat hybridization experiments, the probability of a random false-positive event occurring twice or three times by chance within our sample set of 95 was calculated (p=0.02089 and p=0.001479, respectively). A detailed description of the false-positive rate calculation is given in the “Material and Methods” section.

Figure  2.
Detection of CNVs. The upper part illustrates a region of CNV at 19p13.2 among four individuals. Each short line represents the average fluorescent intensity ratio between sample and reference DNA for an individual BAC clone spotted on the array. The ...

Second, we examined the amount of overlap with previously reported CNVs28,24 (fig. 3). To facilitate the comparison of our CNVs with previously reported CNVs, the locations of all published CNVs were anchored to the same human genome assembly and were mapped to elements in our clone set. As the minimum recurrence of our CNVs increased, so did the proportion that overlapped with previously reported CNVs (fig. 3). Below a recurrence of 3, little overlap was seen between our study and previous studies. This is likely because of false-positive events or very rare CNVs. Between recurrences of 3 and 30, a steadily increasing overlap with previous studies was observed. This may reflect that the more frequent the CNV in the population, the more likely it will be observed in any given study. Beyond a recurrence of 30, no significant increase in overlap was observed. This may reflect the differences in the composition of each study’s population.

Figure  3.
Distribution of overlapped CNVs at different recurrence levels. The percentage of our CNV loci that overlapped with previously reported CNVs were plotted against minimum recurrence levels of CNVs from 1 to 50 within our sample set of 95.

Twenty of the 95 experiments were repeated using fluorochrome reversal. In both the original and the repeat experiments, 771 CNV calls were observed. Of the repeated calls, 81% appeared at least three times in the original CNV discovery sample set of 95. This observation increased confidence for CNVs that were detected three or more times within our sample set. qPCR was performed as a quality check on a small number of loci but was not used for large-scale validation because of the limited throughput of single-locus analysis (see the tab-delimited ASCII file of data set 3 [online only]). For further verification of our calls, five separate hybridizations were repeated using a custom-designed oligonucleotide array covering our 3,654 loci with 389,027 elements (~2 kb spacing between elements) (see the “Material and Methods” section). In the five experiments, 265 CNV calls were confirmed by the oligonucleotide array analysis. Of these CNV calls, 83% were among CNVs detected three or more times in the original CNV discovery set of 95.

We next assessed whether our CNVs coincided with segmental duplications in the genome. To achieve this, we evaluated the segmental-duplication content of the CNVs detected in this study, comparing it against both human and chimpanzee sequences, since there is a significant correlation between contemporary human genome structural variation and historical segmental duplications68 (fig. 4). As the frequency of the CNV increased, so did the enrichment with segmental duplication. This trend increased confidence for CNVs that were observed three or more times in this sample set. We calculated a 5.7-fold duplication enrichment for the most common variants ([gt-or-equal, slanted]5 occurrences in the 95 individuals), which is similar to previous estimates.7,8 Interestingly, the effect was most dramatic (a 12.1-fold increase) for duplications that arose specifically within human.19 In contrast, no enrichment was observed among chimpanzee-only segmental duplications (fig. 4). Elsewhere, we reported an apparent asymmetry with respect to deletion and de novo duplication; 65% of duplications found only in chimpanzee appeared to arise as the result of de novo duplication in the human lineage, as opposed to deletion of a shared duplication in a common ancestor of human and chimpanzee.19 As a result, chimpanzee-only duplications were not expected to be polymorphic in the human lineage.

Figure  4.
Overlap of CNVs with segmental duplications (SD). The percentage of BACs that contain segmental duplications (>10 kb) is graphed against the frequency of the CNV (0 = no variation) for two measures of human segmental duplication (WSSD and WGAC; ...

We also used clustering analysis to assess our CNV calls. We identified the CNVs present within a CEPH family. Clustering of these samples in combination with our original data set samples showed clear grouping of the CEPH family (fig. 5).

Figure  5.
Cluster analysis by use of a CEPH pedigree. Clustering of 105 individuals was based on the high-frequency CNV clones. The 14 CEPH pedigree members are indicated by triangles.

The results from the multiple approaches described above collectively support the presence of novel CNV loci. In addition, the overlaps with previously reported CNVs and segmental duplications, the repeated CNV calls from replicate BAC array CGH experiments and oligonucleotide array hybridizations, the clustering of related individuals on the basis of their CNVs, and the qPCR verification of CNV loci sampled further support their existence. However, the prevalence of these CNVs in the human population can be confirmed only by their presence in multiple individuals. We placed the highest level of confidence in their prevalence when multiple occurrences were observed—for example, 800 loci appeared three or more times in our sample set of 95 individuals. We do not rule out the possibility that true CNVs exist among the loci that we observed at only single and double occurrences in our sample set, since they may represent infrequent events, and a larger sample size will be required to confirm their frequency in the population.

We focused on the high-frequency CNVs (i.e., those found in at least 3 of 95 individuals) for further analysis. There were a total of 9,848 high-frequency CNVs annotated in the 95 individuals analyzed, averaging ~104 per individual. These represent 800 unique loci in the human genome (fig. 6). Strikingly, when these 800 loci are compared with known CNVs, 23% overlap with previously reported CNVs and 77% are novel. The genomic distribution of the 800 CNVs showed no apparent correlation with GC content, imprinted regions, recombination rates, or gene density. Nonrandom somatic alterations—such as the three CNVs associated with immunoglobulin gene rearrangement at chromosomal subbands 2p11.2, 14q32.33, and 22q11.22 (fig. 7)—were detected and removed from further analysis, whereas random somatic alterations not reflecting germline status are not expected to appear recurrently.

Figure  6.
Distribution of CNV clones. High-frequency CNV clones are shown as dots to the right of each chromosome; red, green, and black dots represent presence in three, four or five, and six or more individuals, respectively. Dots to the left of the chromosomes ...
Figure  7.
Detection of immunoglobulin variations. The three parts illustrate expected CNVs associated with the immunoglobulin loci at 2p11.2, 14q32.33, and 22q11.22 (top, middle, and bottom, respectively). The left and right vertical lines represent the average ...

Genomic Diversity within the Sample Population

We next examined the genomic diversity within our sample set. The 800 high-frequency CNV loci (or 1,005 BAC clones) were calculated to span a minimum of 40 Mb of DNA (calculated on the basis of BAC array CGH minimum detection sensitivity of 40 kb per clone15). This equates to ~1.5% of the mapped human autosomes25 that were able to withstand CNV within our sample set. This did not take into account the percentage of single- and double-occurrence loci that represented true CNVs. The two most diverse samples were S73 and S83. They differed at 266 of the high-frequency CNV loci. Then, we asked the question, What is the greatest difference in genome size between two samples within our set? S55 has the highest net gain of CNV clones, at 97, whereas S83 has the highest net loss of CNV clone, at −131. Comparison of these genomes revealed a difference of 228 clones, representing a difference of at least 9 Mb in genomic size between these two individuals.

CNV-Associated Genes

We next identified candidate genes whose dosage may be affected by the 800 CNV loci (fig. 6 and the tab-delimited ASCII file of data set 2 [online only]). In total, 1,673 RefSeq-annotated genes overlapped 546 of the 800 CNV loci. First, we looked for the CNV containing the AMY1A-AMY2A (MIM 104700; MIM 104650) amylase locus, which was a frequently observed copy-number polymorphism.5 This clone was found to be gained in seven individuals and to be lost in five individuals in our sample set (see the tab-delimited ASCII file of data set 2 [online only]). Intriguingly, many genes possibly involved in the senses were found to associate with our CNVs, including a large group of olfactory receptor genes (table 3). In fact, the CNVs associated with olfactory receptor loci segregated in a Mendelian manner in the CEPH family (fig. 8). We also observed genes associated with taste (TAS2R and TAS1R1 [MIM 606225], encoding taste receptors), hearing (ACTG1 [MIM 102560] and MYH9 [MIM 160775]), and sight (OPN1SW [MIM 190900], encoding the short-wave–sensitive cone pigment; GNAT1 [MIM 139330], related to night blindness; and FSCN2 [MIM 607643], IMPDH1 [MIM 146690], and ROM1 [MIM 180721], linked to retinitis pigmentosa) (table 3). In addition, the genes encoding rhesus blood group and defensins were also observed within these common CNVs (see the tab-delimited ASCII file of data set 2 [online only]).

Figure  8.
Inheritance of CNVs at five olfactory receptor loci in 14 members of a CEPH pedigree. The five loci (and clones), in the order shown, are OR2A1 (RP11-466J6), OR2Z1 (RP11-367L15 and RP11-282G19), OR4K1 (RP11-449I24 and CTD-2024K23), OR4M1 (RP11-597A11), ...
Table 3.
Sensory-Related Genes Associated with CNVs

Surprisingly, many genes associated with disease and susceptibility to disease were also found to have CNV among our sample population. For example, a 630-kb region on chromosome 3p21.3 shown to be deleted in lung cancer was observed to be associated with copy-number loss in 20 individuals in this study.26 This region encompasses the putative tumor-suppressor genes TUSC2 (MIM 607052), TUSC4 (MIM 607072), and NAT6 (MIM 607073) (fig. 6, table 4, and the tab-delimited ASCII file of data set 2 [online only]). Many other putative oncogenes and tumor-suppressor genes were also associated with CNVs, such as the VAV2 (MIM 600428) oncogene; RAB3B (MIM 179510), of the RAS oncogene family; TNFRSF25 (MIM 603366); and CDKN1C (MIM 600856) (table 4 and the tab-delimited ASCII file of data set 2 [online only]). In addition to cancer-related genes, CNVs also overlapped genes associated with a bleeding disorder (TBXA2R [MIM 188070]), diabetes mellitus (GCK [MIM 138079]), and spinal muscular atrophy (BSCL2 [MIM 606158], SMA3 [MIM 253400], SMA4 [MIM 271150], and SMN1 [MIM 600354]), as well as with susceptibility to Alzheimer disease (A2M [MIM 103950]), coronary artery disease (LPA [MIM 152200]), and schizophrenia (COMT [MIM 116790]) (table 5). Furthermore, we found 21 human microRNAs that reside within 14 of the high-frequency CNV loci (fig. 6 and table 6).

Table 4.
Select Examples of CNVs Associated with Cancer-Related Genes
Table 5.
Select CNVs Overlapping Genes Associated with Diseases or Disease Susceptibility
Table 6.
MicroRNAs Overlapping CNVs


The existence of large segmental duplications and deletions in the human genome has long been observed through conventional cytogenetic analyses that use light microscopy.27 More recent genomewide analyses with increased resolutions have revealed that CNVs are present throughout the entire human genome26; however, limited genomic coverage of the arrays or the limitations of the various techniques has restricted the discovery of CNVs present in the sample populations. It is currently hypothesized that several thousand CNVs exist within the human genome and thus that most are yet to be discovered.9,28 Here, we used a whole-genome tiling BAC array CGH approach and identified both segmental gains and segmental losses throughout the entire human genome. With complete genome coverage and the tiling nature of our array, we were able to identify a large number of candidate CNVs (3,654). With a focus on only the 800 frequently occurring loci, this study has significantly expanded our knowledge of CNVs. A large proportion (77%) of these high-frequency CNVs are novel; the lack of complete overlap between our CNVs and previously reported CNVs is consistent with the current hypothesis that thousands of CNVs exist in the human population.

In our data set, the net difference in genomic size between two individuals could vary widely, by at least 9 Mb in the two most diverse, representing a difference of 228 distinct CNV clones. In addition, pairwise comparison of the high-frequency CNVs among the 95 individuals revealed that the genomes of the two most diverse individuals differed at 266 loci. These data demonstrate that a significant fraction of the human genome can vary in copy number. On the basis of our high-frequency CNV data set and a minimum detection sensitivity for BAC array CGH of 40 kb, at least 1.5% of the mapped human autosomes is tolerant to CNV. This is an underestimate because the percentage of single- and double-occurrence loci that may represent true CNVs was not taken into account.

Over 1,500 genes were found to overlap the high-frequency CNVs detected in this study. Several of these CNV-associated genes are related to the senses, including a group of olfactory receptor genes, multiple taste-receptor genes, and several genes related to sight or hearing. Genes that are well-known to have variable copy number—such as those encoding rhesus blood group, amylases, and defensins—were also observed within our common CNVs. These associations suggest that CNVs may contribute to phenotypic diversity in humans. Elsewhere, segmental copy-number gains or losses have been demonstrated to associate with developmental disorders and susceptibility to human disease.10 Many genes associated with disease and susceptibility to disease were found to show CNV among the individuals within our study. These include genes associated with diabetes mellitus or a bleeding disorder; cancer-related genes, such as putative oncogenes and tumor-suppressor genes; and genes associated with susceptibility to coronary artery disease or Alzheimer disease. Like other aspects of human genetic variation, understanding of CNVs is critical for studying disease-associated changes correctly, as illustrated in the genome profiling of patients with mental retardation.24 Clinically relevant alterations in copy number need to be separated from a baseline of CNVs for gene discovery. Therefore, it is of utmost importance when genetic association studies of diseases are conducted that they be interpreted in the context of baseline segmental copy-number status; CNVs identified in this study provide a source of information for such a baseline. Interestingly, several of our CNV loci were also found to overlap with microRNAs. Although the functions of microRNAs are largely unknown, they may play a role in the regulation of various biological processes, such as the control of development, differentiation, cell proliferation, and apoptosis, and they have also been linked to human diseases.2931 Recent studies have shown a global downregulation of microRNAs in tumors compared with in normal tissues and an upregulation of microRNA expression via copy-number changes in lymphoma.32,33 Our data raise the possibility that CNVs encompassing microRNAs contribute to human diversity and disease susceptibility.

This comprehensive whole-genome study, identifying both segmental gains and losses in the human population, has significantly expanded our knowledge of CNVs. Remarkably, the genomes of the two most diverse individuals within this study differed by at least 9 Mb in size, or 266 loci in content. In addition, on the basis of our high-frequency CNV data set, at least 1.5% of the human genome is tolerant of CNV. However, with the lack of complete overlap between our CNVs and those identified elsewhere and the hypothesis that thousands of CNVs exist in the human genome, this comprehensive study is still an early step toward a more complete understanding of CNVs within the human population, and more studies are needed to examine the functional roles of CNVs.

Supplementary Material

Data Set 1:
Data Set 2:
Data Set 3:


We thank Media Farshchi and Wendy Peng for computational analysis, Andy Lam and Eric Lee for technical assistance, Sharon Gee for sample collection, the Lam Lab array CGH group for array production, Drs. Carlos Alvarez and Ford Doolittle and members of the Lam Lab for helpful discussions, and especially all donors. This work was supported by funds from Genome Canada/British Columbia, the Canadian Institute of Health Research (to W.L.L. and C.J.B.), the National Institutes of Health (NIH) National Institute of Dental and Cranial Research (to W.L.L.), a Michael Smith Foundation for Health Research scholarship (to R.J.D.), a National Sciences and Engineering Research Council of Canada scholarship (to R.J.D.), and an NIH grant (to E.E.E.). E.E.E. is an Investigator of the Howard Hughes Medical Institute.

Appendix A

Figure A1.

An external file that holds a picture, illustration, etc.
Object name is AJHGv80p91fg9.jpg

Flowchart for calculations. A, Determination of false-positive and false-negative rates in this study by use of six repeat experiments of single female DNA vs male reference DNA, analyzed using our CNV algorithm. B, Calculation for CNV overlaps in replicate experiments.

Web Resources

The URLs for data presented herein are as follows:

BACPAC Resources, http://bacpac.chori.org/genomicRearrays.php(for UCSC May 2004 mapping annotations)
Database of Genomic Variants, http://projects.tcag.ca/variation/
Eisen Lab: Software, http://rana.lbl.gov/EisenSoftware.htm(for Cluster and Treeview)
Gene Expression Omnibus (GEO), http://www.ncbi.nlm.nih.gov/geo/
OMIM, http://www.ncbi.nlm.nih.gov/Omim/ (for BRCA1, BRCA2, APC, MSH2, MSH6, MLH1, AMY1A, AMY2A, TAS1R1, ACTG1, MYH9, OPN1SW, GNAT1, FSCN2, IMPDH1, ROM1,TUSC2, TUSC4, NAT6, VAV2, RAB3B, TNFRSF25, CDKN1C, TBXA2R, GCK, BSCL2, SMA3, SMA4, SMN1, A2M, LPA, and COMT)
Segmental Duplication Database, http://humanparalogy.gs.washington.edu.
UCSC Genome Bioinformatics, http://genome.ucsc.edu/(for May 2004 assembly)
UCSC Human Genome Browser, http://genome.ucsc.edu/cgi-bin/hgGateway.


1. Altshuler D, Brooks LD, Chakravarti A, Collins FS, Daly MJ, Donnelly P (2005) A haplotype map of the human genome. Nature 437:1299–1320. [PMC free article] [PubMed] [Cross Ref]10.1038/nature04226
2. Conrad DF, Andrews TD, Carter NP, Hurles ME, Pritchard JK (2006) A high-resolution survey of deletion polymorphism in the human genome. Nat Genet 38:75–81. [PubMed] [Cross Ref]10.1038/ng1697
3. Hinds DA, Kloek AP, Jen M, Chen X, Frazer KA (2006) Common deletions and SNPs are in linkage disequilibrium in the human genome. Nat Genet 38:82–85. [PubMed] [Cross Ref]10.1038/ng1695
4. McCarroll SA, Hadnott TN, Perry GH, Sabeti PC, Zody MC, Barrett JC, Dallaire S, Gabriel SB, Lee C, Daly MJ, et al (2006) Common deletion polymorphisms in the human genome. Nat Genet 38:86–92. [PubMed] [Cross Ref]10.1038/ng1696
5. Iafrate AJ, Feuk L, Rivera MN, Listewnik ML, Donahoe PK, Qi Y, Scherer SW, Lee C (2004) Detection of large-scale variation in the human genome. Nat Genet 36:949–951. [PubMed] [Cross Ref]10.1038/ng1416
6. Sebat J, Lakshmi B, Troge J, Alexander J, Young J, Lundin P, Maner S, Massa H, Walker M, Chi M, et al (2004) Large-scale copy number polymorphism in the human genome. Science 305:525–528. [PubMed] [Cross Ref]10.1126/science.1098918
7. Sharp AJ, Locke DP, McGrath SD, Cheng Z, Bailey JA, Vallente RU, Pertz LM, Clark RA, Schwartz S, Segraves R, et al (2005) Segmental duplications and copy-number variation in the human genome. Am J Hum Genet 77:78–88. [PMC free article] [PubMed]
8. Tuzun E, Sharp AJ, Bailey JA, Kaul R, Morrison VA, Pertz LM, Haugen E, Hayden H, Albertson D, Pinkel D, et al (2005) Fine-scale structural variation of the human genome. Nat Genet 37:727–732. [PubMed] [Cross Ref]10.1038/ng1562
9. Eichler EE (2006) Widening the spectrum of human genetic variation. Nat Genet 38:9–11. [PubMed] [Cross Ref]10.1038/ng0106-9
10. Inoue K, Lupski JR (2002) Molecular mechanisms for genomic disorders. Annu Rev Genomics Hum Genet 3:199–242. [PubMed] [Cross Ref]10.1146/annurev.genom.3.032802.120023
11. Ishkanian AS, Malloff CA, Watson SK, DeLeeuw RJ, Chi B, Coe BP, Snijders A, Albertson DG, Pinkel D, Marra MA, et al (2004) A tiling resolution DNA microarray with complete coverage of the human genome. Nat Genet 36:299–303. [PubMed] [Cross Ref]10.1038/ng1307
12. Khojasteh M, Lam WL, Ward RK, MacAulay C (2005) A stepwise framework for the normalization of array CGH data. BMC Bioinformatics 6:274. [PMC free article] [PubMed] [Cross Ref]10.1186/1471-2105-6-274
13. Chi B, deLeeuw RJ, Coe BP, MacAulay C, Lam WL (2004) SeeGH—a software tool for visualization of whole genome array comparative genomic hybridization data. BMC Bioinformatics 5:13. [PMC free article] [PubMed] [Cross Ref]10.1186/1471-2105-5-13
14. Locke DP, Sharp AJ, McCarroll SA, McGrath SD, Newman TL, Cheng Z, Schwartz S, Albertson DG, Pinkel D, Altshuler DM, et al (2006) Linkage disequilibrium and heritability of copy-number polymorphisms within duplicated regions of the human genome. Am J Hum Genet 79:275–290. [PMC free article] [PubMed]
15. Pinkel D, Segraves R, Sudar D, Clark S, Poole I, Kowbel D, Collins C, Kuo WL, Chen C, Zhai Y, et al (1998) High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet 20:207–211. [PubMed] [Cross Ref]10.1038/2524
16. Griffiths-Jones S (2004) The microRNA Registry. Nucleic Acids Res 32:D109–D111. [PMC free article] [PubMed] [Cross Ref]10.1093/nar/gkh023
17. Bailey JA, Gu Z, Clark RA, Reinert K, Samonte RV, Schwartz S, Adams MD, Myers EW, Li PW, Eichler EE (2002) Recent segmental duplications in the human genome. Science 297:1003–1007. [PubMed] [Cross Ref]10.1126/science.1072047
18. Bailey JA, Yavor AM, Massa HF, Trask BJ, Eichler EE (2001) Segmental duplications: organization and impact within the current human genome project assembly. Genome Res 11:1005–1017. [PMC free article] [PubMed] [Cross Ref]10.1101/gr.GR-1871R
19. Cheng Z, Ventura M, She X, Khaitovich P, Graves T, Osoegawa K, Church D, DeJong P, Wilson RK, Paabo S, et al (2005) A genome-wide comparison of recent chimpanzee and human segmental duplications. Nature 437:88–93. [PubMed] [Cross Ref]10.1038/nature04000
20. She X, Jiang Z, Clark RA, Liu G, Cheng Z, Tuzun E, Church DM, Sutton G, Halpern AL, Eichler EE (2004) Shotgun sequence assembly and recent segmental duplications within the human genome. Nature 431:927–930. [PubMed] [Cross Ref]10.1038/nature03062
21. Eisen MB, Spellman PT, Brown PO, Botstein D (1998) Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 95:14863–14868. [PMC free article] [PubMed] [Cross Ref]10.1073/pnas.95.25.14863
22. Rozen S, Skaletsky H (2000) Primer3 on the WWW for general users and for biologist programmers. Methods Mol Biol 132:365–386. [PubMed]
23. Weksberg R, Hughes S, Moldovan L, Bassett AS, Chow EW, Squire JA (2005) A method for accurate detection of genomic microdeletions using real-time quantitative PCR. BMC Genomics 6:180. [PMC free article] [PubMed] [Cross Ref]10.1186/1471-2164-6-180
24. de Vries BB, Pfundt R, Leisink M, Koolen DA, Vissers LE, Janssen IM, Reijmersdal S, Nillesen WM, Huys EH, Leeuw N, et al (2005) Diagnostic genome profiling in mental retardation. Am J Hum Genet 77:606–616. [PMC free article] [PubMed]
25. International Human Genome Sequencing Consortium (2004) Finishing the euchromatic sequence of the human genome. Nature 431:931–945. [PubMed] [Cross Ref]10.1038/nature03001
26. Lerman MI, Minna JD, for The International Lung Cancer Chromosome 3p21.3 Tumor Suppressor Gene Consortium (2000) The 630-kb lung cancer homozygous deletion region on human chromosome 3p21.3: identification and evaluation of the resident candidate tumor suppressor genes. Cancer Res 60:6116–6133. [PubMed]
27. Seabright M (1971) A rapid banding technique for human chromosomes. Lancet 2:971–972. [PubMed] [Cross Ref]10.1016/S0140-6736(71)90287-X
28. Lee C (2005) Vive la difference! Nat Genet 37:660–661. [PubMed] [Cross Ref]10.1038/ng0705-660
29. Alvarez-Garcia I, Miska EA (2005) MicroRNA functions in animal development and human disease. Development 132:4653–4662. [PubMed] [Cross Ref]10.1242/dev.02073
30. Bartel DP (2004) MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 116:281–297. [PubMed] [Cross Ref]10.1016/S0092-8674(04)00045-5
31. Wienholds E, Plasterk RH (2005) MicroRNA function in animal development. FEBS Lett 579:5911–5922. [PubMed] [Cross Ref]10.1016/j.febslet.2005.07.070
32. Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, Sweet-Cordero A, Ebert BL, Mak RH, Ferrando AA, et al (2005) MicroRNA expression profiles classify human cancers. Nature 435:834–838. [PubMed] [Cross Ref]10.1038/nature03702
33. Tagawa H, Seto M (2005) A microRNA cluster as a target of genomic amplification in malignant lymphoma. Leukemia 19:2013–2016. [PubMed] [Cross Ref]10.1038/sj.leu.2403942

Articles from American Journal of Human Genetics are provided here courtesy of American Society of Human Genetics


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Cited in Books
    Cited in Books
    PubMed Central articles cited in books
  • GEO DataSets
    GEO DataSets
    GEO DataSet links
  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...