Detection of Structural Variants (SVs) and Copy Number Variations (CNVs) on NGS Data

1
Detection of Structural Variants (SVs) and
Copy Number Variations (CNVs) on NGS Data
2
SVs and CNVs

• They are often confused
• SVs regions contain insertions and deletions
  (indels) or inversions.
• CNVs regions appearing a different number of
  times in different individuals.
• They are closely related phenomena.
• SVs are operationally defined as genomic events
  involving gt50bp. They include CNVs as well as
  rearrangements such as inversions and
  translocations.

3
Why studying SVs and CNVs

• It has been acknowledged only very recently that
  human genomes differ more as a consequence of SVs
  (including CNVs) than of single-base differences.
  first "hypothesis" in 2004-2005 not taken
  seriously, and evidence only in 2010 with NGS.
• In particular, many studies observed CNVs and did
  genotyping with them they are the easiest SVs to
  detect.
• I am not aware of tools for detecting inversions
  and translocations of DNA durectly on NGS data
• There are some tools for detecting CNVs from now
  on we discuss them only.

4
Copy Number Variations

• The challenge is to discover effects of CNVs on
  human diseases, complex traits (combination of
  genetic and environmental effects), and
  evolution.
• Genotyping of human CNVs is far from being a
  routine procedure. This is a limit to
  personalized medicine, for which CNVs detection
  is a crucial step.
• No standard method (many and very recent tools,
  not yet a fair/sharp comparison) exists.

5
Detecting CNV

• Since late nineties and until very recently (and
  still) CNVs were/are detected using aCGH Array
  Comparative Hybridization.
• The array platform are not as rapid and cheap as
  NGS, and their data cannot be re-used once
  processed.
• The aGCH Array has inherent limits on the size
  and frequency of detectable CNVs.
• NGS opened a new era in CNV-detection!

6
CNV calling

• There are three approaches for CNV calling
• Based on read count (RC), or read alignment
  coverage (Breakdancer, CNVnator, CNV-seq, and
  tools of Campbell et al, 2008, Alkan et al,
  2009, Sudmant et al, 2010, Yoon et al,
  2009).
• Based on paired end reads (PEMer, CNVer,
  VariationHunter, MoDIL, Breakdancer, tools of
  Sindi et al, 2009, Quinlan et al, 2010).
• Based on split-read alignments (Pindel, Mosaik,
  tool of Mills et al, 2006).

7
Read Count Approaches

• Measuring the amount of reads mapped to a
  location in the reference genome.
• Identify CNV regions
• Estimating Copy Number
• Some use a sliding window.
• Problems when coverage is not uniform within a
  CNV

8
Paired End Reads Approaches

• Mate/Paired End Reads are mapped on the reference
  genome.

9
Split Read Approach

• A read is not mapped in a single locations
  because of possible structural variation.
• All un-aligned reads are split and then mapping
  is sought again.
• Iterate to find the actual breakpoint.