Genome structural variation discovery and genotyping

Corresponding author: Evan Eichler
Principal Investigator
Professor of Genome Sciences
Dept. Genome Sciences, U. Washington, Seattle, WA

Key points

Structural variation was originally defined as insertions, deletions and inversions greater than 1 kb in size, but with the sequencing of human genomes now becoming routine, the operational spectrum of structural variants has widened to include events >50 bp in length.
The main focus of structural variant (SV) studies should be accurate characterization of the copy, content and structure of genomic variants.
Methods to discover and genotype structural variation can be divided into two main types: experimental and computational.
Experimental methods for discovering SVs include hybridization-based approaches (SNP microarrays and array comparative genomic hybridization) and single-molecule analysis (optical mapping). In addition, PCR-based techniques can be used to genotype SVs.
Computational methods use genome sequencing data to discover and genotype SVs. There are four main computational approaches: read-pair, read-depth, split-read and sequence-assembly methods.
All existing platforms and methods have different biases and limitations. Accurate characterization of the full spectrum of structural variation remains a challenge.

1. Associations between disease and structural variation

large variants (typically gains and losses several hundred kilobase pairs in length) that are individually rare in the population (<1%) but collectively account for a significant fraction of disease, as seen for some neurological and neurocognitive disorders.
multicopy gene families that are commonly copy number variable and contribute to disease susceptibility, as seen for traits related to immune gene functions.

2. Classes of structural variation

Classes of structural variation

3. Methods

3.1 Hybridization-based microarray approaches

Both hybridization-based technologies infer copy number gains or losses compared to a reference sample or population

Array CGH.
原理：将等量的待测DNA和正常对照DNA分别用红色和绿色荧光染料标记，混合，然后与全基因组DNA芯片进行竞争性杂交。杂交后的芯片经激光扫描，比较每个点红光和绿光的发光强度。若红光过强，表明待测样本拷贝数复制；若红光较弱，表明待测样本拷贝数缺失；若红绿光均等，则表明待测样品拷贝数正常。
应用：可以检测全基因组水平上的CNV。
点评：aCGH的分辨率，取决于芯片中探针在全基因组中的密度和探针长度。安捷伦SurePrint G3 Human CGH Microarray 1×1M芯片中，探针间距约为2kb。因为来自一个点的荧光的光强变化，可能会带有一定的偶然性，所以，一般是看染色体空间位置上相邻的3个点（或者更多的点），如果这3个点的荧光比值，都发生同一个方向的偏离，就可以判断这一段有拷贝数变异的证据。基于这点考虑，按照3个点计算，aCGH的分辨率约为6kb。
SNP arrays.
原理： 与aCGH采用的双杂交策略不同的是，SNP-array利用待测样本与芯片探针进行单杂交，通过比较不同样本信号的强度来确定每个位点的拷贝数。
应用：SNP-array芯片的探针为SNP位点序列，可以提供SNP信息。除可检测CNV外，还可检测单亲二倍体(UPD)、杂合性缺失(LOH)和嵌合体。
点评：SNP 芯片探针在全基因组上的的密度非常大，分辨率很高。但是这些探针在基因组中并非均衡分布，在一些重复序列和复杂的CNV 区域， SNP 密度是较小的，不能得到较为清晰的CNV 图谱。Affymetrix 公司和Illumina 公司在新一代芯片中增加一个非多态性的探针，将探针更好地定位在特定区域，提高图谱的清晰度。目前Affymetrix 公司的CytoScan HD芯片为探针密度最高的芯片，含有270万个探针，基因间探针间距平均为1kb。按照3个点计算，SNP-array的分辨率约为3kb。
局限
1.array限于检测用于设计探针中存在的序列的拷贝数差异，不提供关于复制拷贝的位置的信息，并且通常不能解决单碱基对水平的断点；
2.检测single-copy gains (3 to 2 copy-number ratio)时的灵敏度小于deletions (1 to 2 copy-number ratio)；
3.由于检出偏倚造成小事件的检出难度更大smaller events detected by array platforms are overwhelmingly deletions, partly owing to an ascertainment bias检出偏倚；
4.同源缺失更容易检出Homozygous deletions are the easiest class to detect regardless of platform and can be detected with fewer probes than single-copy gains and losses;
5.不同公司平台结果需要不同的算法解读；
6.重复序列区域无法应用 the most important limitation of arrays is the use of hybridization-based assays in repeat-rich and duplicated regions. Array CGH and SNP platforms assume each location to be diploid in the reference genome, which is not valid in duplicated sequence.
优点
通量和成本 throughput and cost

3.2 Single-molecule analysis

Approaches such as fluorescent in situ hybridization (FISH), fiber-FISH and spectral karyotyping（核型分析） provided our first glimpses of common and rare genome structural variation. However, their low throughput and low resolution limit their application to a few individuals and to particularly large structural differences (~500 kb to 5 Mb)

3.3 Sequencing-based computational approaches

算法挑战。

四种检测SV的算法

read-pair (RP)
read-depth：只能检测插入、缺失 can be used to detect only losses (deletions) and gains (duplications), and cannot discriminate between tandem and interspersed duplications；
split-read：通过分析读序列与参考基因组的序列比对，能够准确地检测出所有变异类的断点，然而，它们通常比其他方法需要更长的读取时间，并且在重复序列中表现较差；
assembly methods

3.4 Genotyping

PCR-based techniques
SNP-array-based techniques
Array CGH-based techniques
Sequencing-based approaches