Copy number Variants (CNVs), which comprise deletions, insertions and inversions of genomic sequence, are a main form of genetic variation between individual genomes. CNVs are commonly present in the genomes of human and other species. However, they have not been extensively characterized as their ascertainment is challenging. I reviewed current CNV studies and CNV discovery methods, especially the algorithms which infer CNVs from whole genome Single Nucleotide Polymorphism (SNP) arrays and compared the performance of three analytical tools in order to identify the best method of CNV identification. Then I applied this method to identify CNV events in three European population isolates—the island of Vis in Croatia, the islands of Orkney in Scotland and villages in the South Tyrol in Italy - from Illumina genome-wide array data with more than 300,000 SNPs. I analyzed and compared CNV features across these three populations, including CNV frequencies, genome distribution, gene content, segmental duplication overlap and GC content. With the pedigree information for each population, I investigated the inheritance and segregation of CNVs in families. I also looked at association between CNVs and quantitative traits measured in the study samples. CNVs were widely found in study samples and reference genomes. Discrepancies were found between sets of CNVs called by different analytical tools. I detected 4016 CNVs in 1964 individuals, out of a total of 2789 participants from the three population isolates, which clustered into 743 copy number variable regions (CNVRs). Features of these CVNRs, including frequency and distribution, were compared and were shown to differ significantly between the Orcadian, South Tyrolean and Dalmatian population samples. Consistent with the inference that this indicated population-specific CNVR identity and origin, it was also demonstrated that CNV variation within each population can be used to measure genetic relatedness. Finally, I discovered that individuals who had extreme values of some metabolic traits possessed rare CNVs which overlapped with known genes more often than in individuals with moderate trait values.

No description available.

No description available.

Introduction: The main risk factor for the development of oesophageal adenocarcinoma is Barrett’s oesophagus (BE). To diagnose those patients who will progress to cancer early to improve the dismal survival rate of oesophageal adenocarcinoma, patients with BE undergo regular endoscopic surveillance. The vast majority of patients, however, will never progress and are therefore monitored unnecessarily. Copy number changes have been shown to be important in the progression of BE to oesophageal adenocarcinoma (Li et al., 2014). Shallow whole genome sequencing (sWGS) has been established as a cost-effective method of investigating copy number changes in formalin fixed paraffin embedded (FFPE) tissue (Scheinin et al., 2014). We hypothesised that copy number alterations may be valuable markers in disease progression and aimed to characterise them in the progression of Barrett’s using sWGS in order to predict progression in patients from a point in time as close to baseline endoscopy as possible and to integrate p53 staining. Methods: To optimise sWGS we compared 50X WGS on frozen tissue with 0.1X WGS from FFPE tumour material from the same patient. To address poor cellularity in endoscopic biopsies, shallow WGS data from a 50% cellularity biopsy with a 90% frozen sample from a single patient were compared. Accounting for poor biopsy cellularity 0.4X coverage was used. We performed FFPE shallow WGS on 806 samples from an 89-patient cohort comprising a 1:1 ratio of patients who progressed to high grade dysplasia (HGD) and patients who never progressed. 1-31 samples per patient were collected over time and space throughout surveillance. Non-progressors had significantly longer follow-up (p-value = 0.0008). Data was processed based on published bioinformatic pipelines. Copy number analysis was carried out using a generalised linear model (GLM) in order to develop a predictive algorithm. Results: During optimisation, ˃85% of copy number changes were detected in both frozen and FFPE samples from spatially distinct regions of an individual tumour. We found 91% and 93% agreement in copy number calls using orthogonal platforms between 90% (frozen) and 50% (FFPE) cellularity samples from one tumour. In the 806 sample Barrett’s cohort, we observed larger copy number alterations in patients who progressed to cancer compared with non-progressors and significantly more CN alterations in progressor patients (p-value ˂ 0.001). More cancer-associated genes were affected in progressors and we observed significant heterogeneity between patients. There was also a greater level of complexity seen in the progressor patients when analysed using affinity propagation clustering. These data allowed us to develop a regression model to predict progression. Using the GLM model, we successfully classified samples as early as progressor or not with an AUC of 85.75% and a sensitivity and specificity of 84 and 79% respectively. At the patient level 94% progressor patients had at least one sample classified as at risk of progression and non-dysplastic progressor samples were classified as early as 13 years prior to HGD diagnosis. Depending on the classification threshold used, all samples over time and space were not classified as being at risk of progression in at least 60% patients who have not yet progressed to HGD/cancer. We observed 2 pathways to progression supporting previous observations. 90% of progressors had samples prior to their HGD or cancer diagnosis classified as being predisposed to progression suggestive of genetically unstable lesions from early on in surveillance that progressed to HGD over time. The remaining 10% appeared as non-progressors until their diagnosis of HGD. We investigated p53 expression in our patient cohort as the only biomarker to have successfully transitioned into the clinic for Barrett’s surveillance. Whilst we found our cohort to be representative in staining compared to other published cohorts, it did not contribute to the GLM and the copy number data out-performed the use of p53 IHC in the context of Barrett’s surveillance. Conclusions: We have optimised the use of shallow WGS in oesophageal adenocarcinoma and Barrett’s. Using these copy number data, we can confidently distinguish between patients who will progress to cancer and the majority of patients who will never progress. This approach has led to the development of a model for predicting progression in the clinical setting which is promising for further clinical validation.

Schizophrenia (OMIM 181500) is an incurable and severe psychiatric disorder comprised of three symptom domains (positive symptoms, negative symptoms and cognitive impairments) with a worldwide prevalence of approximately 1%. There is a substantial amount of evidence demonstrating that schizophrenia has a strong a genetic component. Broad-sense heritability estimates range from 64-80% and first-degree relatives of schizophrenia patients have 10-fold increased risk of developing the disorder compared to the general population. It is thought that both single nucleotide polymorphisms and copy number variants (CNVs) contribute to the heritability of schizophrenia. This thesis focuses on the role of CNVs in the etiology of schizophrenia. We performed a genome-wide CNV analysis of 166 schizophrenia patients and 52 psychiatrically healthy controls. In our overall CNV analysis we did not find any significant differences between cases and controls across a variety of CNV categories, nor did we find significant differences when CNVs were partitioned by size (small, medium or large). However, we were the first group to consider small CNVs (< 100-500 kb) in a multiple-hit model where we observed that a slightly higher proportion of case subjects had two-or-more conservative CNVs. We defined a CNV as conservative if it met any of the following three criteria: 1) a known deleterious CNV, 2) a CNV > 1 Mb that was novel to the Database of Genomic Variants (DGV) or 3) a CNV < 1 Mb that was novel to the DGV and that overlapped the coding region of a gene of interest. Genes of interest included genes with a previous association with a neuropsychiatric disorder, or genes with high or specific brain expression, or an association with any other neurocognitive or neuropsychiatric disorders. Two of our case subjects who harbored the highest amount of conservative CNVs also shared a 15q11.2 breakpoint 1-2 (BP1-2) deletion which is a compelling candidate risk locus for schizophrenia. We also found that a slightly higher proportion of case subjects harbored clinically significant CNVs (conservative CNVs > 1 Mb or clinically recognized as deleterious) when compared to controls. Additionally, we hypothesized that individuals with more severe CNVs would show more neurocognitive deficits and more pronounced abnormalities in brain structure volume, however, we had largely negative results. We also reported a case of childhood-onset schizophrenia who had three large chromosomal abnormalities including a paternally inherited 2.2 Mb deletion of chromosome 3p12.2-p12.1, a de novo 17.6 Mb duplication of chromosome 16q22.3-q24.3 and a de novo 43 Mb deletion of chromosome Xq23-q28. We were able to confirm previous reports of CNV findings in schizophrenia such as the involvement of large, rare and de novo CNVs. In addition, the work in this thesis leads us to propose a multiple-hit CNV model which requires a shift in the way we currently approach schizophrenia genetics. First, we must identify all CNVs, especially those of smaller size (< 100 kb). Next, we require a more precise understanding of the impact that CNVs have on gene expression, especially in the brain. With all of the right tools in place, we can move towards a disease model for schizophrenia that considers the totality of CNVs in any given individual. We propose that the use of recurrent CNVs such as the 15q11.2 BP1-2 CNV is a good starting point for studying a multiple-hit CNV model.

DNA copy number alterations (CNAs) are genetic changes that can produce adverse effects in numerous human diseases, including cancer. CNAs are segments of DNA that have been deleted or amplified and can range in size from one kilobases to whole chromosome arms. Development of array comparative genomic hybridization (aCGH) technology enables CNAs to be measured at sub-megabase resolution using tens of thousands of probes. However, aCGH data are noisy and result in continuous valued measurements of the discrete CNAs. Consequently, the data must be processed through algorithmic and statistical techniques in order to derive meaningful biological insights. We introduce model-based approaches to analysis of aCGH data and develop state-of-the-art solutions to three distinct analytical problems. In the simplest scenario, the task is to infer CNAs from a single aCGH experiment. We apply a hidden Markov model (HMM) to accurately identify CNAs from aCGH data. We show that borrowing statistical strength across chromosomes and explicitly modeling outliers in the data, improves on baseline models. In the second scenario, we wish to identify recurrent CNAs in a set of aCGH data derived from a patient cohort. These are locations in the genome altered in many patients, providing evidence for CNAs that may be playing important molecular roles in the disease. We develop a novel hierarchical HMM profiling method that explicitly models both statistical and biological noise in the data and is capable of producing a representative profile for a set of aCGH experiments. We demonstrate that our method is more accurate than simpler baselines on synthetic data, and show our model produces output that is more interpretable than other methods. Finally, we develop a model based clustering framework to stratify a patient cohort, expected to be composed of a fixed set of molecular subtypes. We introduce a model that jointly infers CNAs, assigns patients to subgroups and infers the profiles that represent each subgroup. We show our model to be more accurate on synthetic data, and show in two patient cohorts how the model discovers putative novel subtypes and clinically relevant subgroups.

DNA copy number alterations (CNAs) are genetic changes that can produce adverse effects in numerous human diseases, including cancer. CNAs are segments of DNA that have been deleted or amplified and can range in size from one kilobases to whole chromosome arms. Development of array comparative genomic hybridization (aCGH) technology enables CNAs to be measured at sub-megabase resolution using tens of thousands of probes. However, aCGH data are noisy and result in continuous valued measurements of the discrete CNAs. Consequently, the data must be processed through algorithmic and statistical techniques in order to derive meaningful biological insights. We introduce model-based approaches to analysis of aCGH data and develop state-of-the-art solutions to three distinct analytical problems. In the simplest scenario, the task is to infer CNAs from a single aCGH experiment. We apply a hidden Markov model (HMM) to accurately identify CNAs from aCGH data. We show that borrowing statistical strength across chromosomes and explicitly modeling outliers in the data, improves on baseline models. In the second scenario, we wish to identify recurrent CNAs in a set of aCGH data derived from a patient cohort. These are locations in the genome altered in many patients, providing evidence for CNAs that may be playing important molecular roles in the disease. We develop a novel hierarchical HMM profiling method that explicitly models both statistical and biological noise in the data and is capable of producing a representative profile for a set of aCGH experiments. We demonstrate that our method is more accurate than simpler baselines on synthetic data, and show our model produces output that is more interpretable than other methods. Finally, we develop a model based clustering framework to stratify a patient cohort, expected to be composed of a fixed set of molecular subtypes. We introduce a model that jointly infers CNAs, assigns patients to subgroups and infers the profiles that represent each subgroup. We show our model to be more accurate on synthetic data, and show in two patient cohorts how the model discovers putative novel subtypes and clinically relevant subgroups.

Epithelial ovarian cancer is the leading cause of death by gynaecological malignancy. Elucidation of the driver genes of ovarian cancer will lead to treatment targets and tailored therapy for this disease. The Affymetrix Genome-Wide SNP Array 6.0 was used to study 100 serous ovarian samples and 10 normal ovarian samples to identify loci and driver genes. The ovarian cancer genome was found to have high overall genomic instability across all chromosomes and key known genes in this disease were identified in the dataset. Aberrant regions of copy number gain were located in “blocks” of constant copy number at 1p, 1q, 8q, 12p, 19q and 20q. The range in copy number for gains was 4.2 to 5.1. The “blocks” of genes were located at 8p and 5p for copy number losses. The range for copy number loss was 0.6 to 0.9.

This study presents an integrated analysis of pancreatic ductal adenocarcinomas (PDACs) for identification of putative cancer driver genes in somatic copy number gains (SCNGs). SCNG data on 60 PDAC genomes was extracted to identify 756 genes, mapping to 20 genomic loci that are recurrently gained. Through copy number and gene expression analysis on a panel of 29 human pancreatic cancer cell lines, this gene catalogue was refined to 34 PDAC high-confidence candidate genes. The performance of these genes was assessed in pooled shRNA screens and only ECT2 showed significant essentiality to cell viability in specific PDAC cell lines with genomic gains at the 3q26.3 locus that harbor this gene. Targeted shRNA-mediated interference of ECT2, as well as pharmacological inhibition, are supportive of the pooled shRNA screen findings. These results favor ECT2 as a candidate target gene for further evaluation in the subset of PDACs presenting with 3q26 somatic copy number gains.

Epithelial ovarian cancer is the leading cause of death by gynaecological malignancy. Elucidation of the driver genes of ovarian cancer will lead to treatment targets and tailored therapy for this disease. The Affymetrix Genome-Wide SNP Array 6.0 was used to study 100 serous ovarian samples and 10 normal ovarian samples to identify loci and driver genes. The ovarian cancer genome was found to have high overall genomic instability across all chromosomes and key known genes in this disease were identified in the dataset. Aberrant regions of copy number gain were located in “blocks” of constant copy number at 1p, 1q, 8q, 12p, 19q and 20q. The range in copy number for gains was 4.2 to 5.1. The “blocks” of genes were located at 8p and 5p for copy number losses. The range for copy number loss was 0.6 to 0.9.