Detecting structural variations in the human genome using next generation sequencing

Abstract

Structural variations are widespread in the human genome and can serve as genetic markers in clinical and evolutionary studies. With the advances in the next-generation sequencing technology, recent methods allow for identification of structural variations with unprecedented resolution and accuracy. They also provide opportunities to discover variants that could not be detected on conventional microarray-based platforms, such as dosage-invariant chromosomal translocations and inversions. In this review, we will describe some of the sequencing-based algorithms for detection of structural variations and discuss the key issues in future development.

Figure 1:

CNV detection using single-end reads with a control (reference) genome. The triangles represent mapped read positions along the case and control genomes. Regions in which the tag counts are different with statistical significance are identified as potential CNVs.

Figure 2:

Configurations of PEMs in various types of SVs. (A) Deletion. The paired-end read spans the breakpoint of a deletion. Thus, the mapped distance of the paired-end reads is significantly larger than the insert size. (B) Insertion. The paired-end reads spans an insertion, and the mapped distance significantly less than the insert size. (C) Inversion. The read pair encompasses one breakpoint of an inversion and the right end is mapped with incorrect orientation. (D) Tandem duplication. The read pair spans the middle breakpoint of a tandem duplication. The PEM will have correct orientation but with reverse order. (E) Intra-chromosomal translocation. Two read pairs span the two breakpoints of an intra-chromosomal translocation with one pair having a large mapped distance and the other having correct orientation but their ordering reversed. (F) Inter-chromosomal translocation. The two ends of the pair are mapped to different chromosomes. (G) One-end unmapped. The sequenced genome has a DNA segment that does not exist in the reference genome. One end of the pair is mappable but the other is not.

Figure 3:

The insert size is larger than the duplicated segment in a tandem duplication. (A) The read pair spans the middle breakpoint. This could be misleading because one may mistakenly conclude that there is an insertion between the pair. (B) The read pair spans one of the duplicated segments.

Figure 4:

Local distribution of mapped distances of the paired-end reads. (A) A homozygous deletion. The distribution is shifted to right, where the gray curve is the global distribution of the mapped distance. (B) A heterozygous deletion. The distribution is a mixture of two distributions and therefore is a bimodal distribution.

Figure 5:

Split mapping. (A) The read that spans the breakpoint of a deletion can be split and mapped to two positions on the reference genome. (B) The read that contains a small insertion is mapped to the reference genome after removing the small insertion.