Computational tools for CNV detection using probe-level analysis of Affymetrix SNP arrays : application to the study of CNVs in follicular lymphoma

• Main Author: Farnoud, Noushin
• Language: English
• Published: University of British Columbia 2012
• Online Access: http://hdl.handle.net/2429/43034

Copy number variants (CNVs) account for both variations among normal individuals and pathogenic variations. The introduction of DNA microarrays had a significant impact on the resolution of detectable CNVs and yielded a new perspective on the submicroscopic CNVs. Oligonucleotide microarrays, such as Affymetrix SNP arrays, have been commonly used for genome-wide CNV analysis. Despite the improvements in the technology, a major concern of using microarrays is how a putative CNV is defined. A disadvantage of oligonucleotide arrays is the poor signal-to-noise ratio of the data that leads to considerable variation in reported intensity readouts. Such variation will lead to false positive and false negative results, regardless of how the data are analysed. The most common approach to circumvent this problem is looking for abrupt ratio intensity shifts in several consecutive markers (e.g., SNP probes). However this approach reduces the overall resolution and mitigates the sensitivity of detecting CNVs with fewer probes. This limitation emphasizes the importance of designing methods that can identify noisy readouts at the probe-level. The main goals of this work were to study the scale of the variability in Affymetrix SNP arrays and to develop computational tools that can improve the resolution of CNV detection. By using simulated data, it was shown that the proposed method improved the accuracy and precision of detecting CNVs with fewer probes compared to standard methods. This approach was also applied to tumor/normal pairs from 25 follicular lymphoma patients and 286 candidate CNVs were found, from which 261 (91.2%) were also seen by other array-based method(s). Importantly, from 32 novel deletions, undetected by other array-based methods, at least 15 (47%) were real based on sequence-based validation. An example of a novel discovery was a partial deletion of the extracellular domain of the KIT proto-oncogene that may lead to constitutive activation of this gene. Gain of function mutations of KIT has been previously reported in several other hematologic cancers through other mechanisms such as point mutations. In conclusion, CNV discovery contributes to our understanding of complex diseases and the methods presented here should provide means for better detection of CNVs and their interpretation.