Miroslaw K. Kapuscinski

Supercomputing enabling exhaustive statistical analysis of genome wide association study data: Preliminary results

Most published GWAS do not examine SNP interactions due to the high computational complexity of computing p-values for the interaction terms. Our aim is to utilize supercomputing resources to apply complex statistical techniques to the worlds accumulating GWAS, epidemiology, survival and pathology data to uncover more information about genetic and environmental risk, biology and aetiology. We performed the Bayesian Posterior Probability test on a pseudo data set with 500,000 single nucleotide polymorphism and 100 samples as proof of principle. We carried out strong scaling simulations on 2 to 4,096 processing cores with factor 2 increments in partition size. On two processing cores, the run time is 317h, i.e. almost two weeks, compared to less than 10 minutes on 4,096 processing cores. The speedup factor is 2,020 that is very close to the theoretical value of 2,048. This work demonstrates the feasibility of performing exhaustive higher order analysis of GWAS studies using independence testing for contingency tables. We are now in a position to employ supercomputers with hundreds of thousands of threads for higher order analysis of GWAS data using complex statistics.

‘Next-generation’ genome wide association studies

The first wave of cancer genome-wide association studies (GWAS) have revealed tens of independent loci marked by common variants of unknown or likely no functional significance that explain about 5-10% of familial risk for the particular disease. The approach taken to date has been conservative, and only a fraction of information has yet to be extracted from these expensive enterprises. For example, the Bonferroni procedure for selecting candidate phase II SNPs ignores many SNPs that happen to fail an extremely low p-value threshold. While this procedure does guarantee control of false positives, it seems counterintuitive to the purpose of phase I, which is to generate hypotheses based on promising candidates. Researchers have generally combined data from the discovery phase I and other phases and used ‘genome-wide thresholds’ based on assuming all SNPs are independent. Linkage disequilibrium (LD) makes it problematic to differentiate a real signal from highly correlated proxy signals. Most published GWAS do not examine SNP interactions due to: (a) the high computational complexity of computing pvalues for the interaction terms, and (b) the typically low power to detect significant interactions. It is plausible that more information should be extracted if: (i) higher order interactions are fitted, (ii) highly selected cases and controls are used in phase I, (iii) large replication studies are used, especially if involving existing GWAS data, (iv) the non-independence of SNPs is taken into account using, e.g. BEAGLE CALL or haplotype analyses, (v) focus is on candidate gene pathways, and/or functional SNPs, and (vi) rarer and more SNPs, such as is available from the Illumina 5M SNP chip, are used. We will illustrate these ideas using data from a GWAS of early-onset breast cancers, enriched for those with a family history, and a GWAS using extremes sample of extremes for mammographic density. We will also discuss the design of a large international breast cancer GWAS using the Illumina 5M SNP chip, phase I cases enriched for family history, population-based phase II cases and controls, population-based family study of candidate SNPs, and GxG analyses using ‘massively parallel’ super computing.

Ability of known susceptibility SNPs to predict colorectal cancer risk for persons with and without a family history

Background A number of single nucleotide polymorphisms (SNPs), which are common inherited genetic variants, have been identified that are associated with risk of colorectal cancer. The aim of this study was to determine the ability of these SNPs to estimate colorectal cancer (CRC) risk for persons with and without a family history of CRC, and the screening implications. Methods We estimated the association with CRC of a 45 SNP-based risk using 1,181 cases and 999 controls, and its correlation (r) with CRC risk predicted from detailed family history. We estimated the predicted change in the distribution across predefined risk categories, and implications for recommended age to commence screening, from adding SNP-based risk to family history. Results The inter-quintile risk ratio for colorectal cancer risk of the SNP-based risk was 2.46 (95% CI 1.91 – 3.11). SNP-based and family history-based risks were not correlated (r = 0.02). For persons with no first-degree relatives with CRC, recommended screening would commence 2 years earlier for women (4 years for men) in the highest quintile of SNP-based risk, and 12 years later for women (7 years for men) in the lowest quintile. For persons with two first-degree relatives with CRC, recommended screening would commence 15 years earlier for men and women in the highest quintile, and 8 years earlier for men and women in the lowest quintile. Conclusions Risk reclassification by 45 SNPs could inform targeted screening for CRC prevention, particularly in clinical genetics settings when mutations in high-risk genes cannot be identified.

Genetic susceptibility markers for a breast-colorectal cancer phenotype: Exploratory results from genome-wide association studies

Background Clustering of breast and colorectal cancer has been observed within some families and cannot be explained by chance or known high-risk mutations in major susceptibility genes. Potential shared genetic susceptibility between breast and colorectal cancer, not explained by high-penetrance genes, has been postulated. We hypothesized that yet undiscovered genetic variants predispose to a breast-colorectal cancer phenotype. Methods To identify variants associated with a breast-colorectal cancer phenotype, we analyzed genome-wide association study (GWAS) data from cases and controls that met the following criteria: cases (n = 985) were women with breast cancer who had one or more first- or second-degree relatives with colorectal cancer, men/women with colorectal cancer who had one or more first- or second-degree relatives with breast cancer, and women diagnosed with both breast and colorectal cancer. Controls (n = 1769), were unrelated, breast and colorectal cancer-free, and age- and sex- frequency-matched to cases. After imputation, 6,220,060 variants were analyzed using the discovery set and variants associated with the breast-colorectal cancer phenotype at P<5.0E-04 (n = 549, at 60 loci) were analyzed for replication (n = 293 cases and 2,103 controls). Results Multiple correlated SNPs in intron 1 of the ROBO1 gene were suggestively associated with the breast-colorectal cancer phenotype in the discovery and replication data (most significant; rs7430339, Pdiscovery = 1.2E-04; rs7429100, Preplication = 2.8E-03). In meta-analysis of the discovery and replication data, the most significant association remained at rs7429100 (P = 1.84E-06). Conclusion The results of this exploratory analysis did not find clear evidence for a susceptibility locus with a pleiotropic effect on hereditary breast and colorectal cancer risk, although the suggestive association of genetic variation in the region of ROBO1, a potential tumor suppressor gene, merits further investigation.

Abstract 1975: The 3D chromatin structure of the PTHLH region comprises a dynamic hierarchical looping complex that approximates the protein-coding genes and facilitates promoter and enhancer promiscuity

Parathyroid Hormone-Like Hormone (PTHLH) is an important regulatory gene encoding the product Parathyroid Hormone-related Protein (PTHrP). Among many biological roles, it has been demonstrated to be essential to the development of multiple tissue-types, the regulation of foetal calcium levels, and the metastasis of breast cancer to bone. Its region has GWAS associations with breast cancer, breast size, height, type 2 diabetes mellitis, neural development, cardiac arrest, and several immunological phenotypes. They extend both up- and down-stream of PTHLH, however their driving molecular mechanisms remain obscure. Analysis of datasets including Hi-C, ChIA-PET, IM-PET and ChIP-seq suggests PTHLH sits within a 1.3Mb Topologically Associating Domain (TAD) containing 4 other protein-coding genes (pc-genes) CCDC91, MRPS35, MANSC4 and KLHL42. PTHLH itself rests between two 220kb inducible sub-TADs that likely comprise regulatory archipelagos directed primarily at its regulation. There are several other inducible loops in the TAD. All protein-coding genes within the TAD feature an Activated Chromatin Hub (ACH) nearby their 5’. The 3D-structure of the TAD appears to be supported by a system of permanent and dynamic tether points producing primary and inducible chromatin loops. These result in all pc-genes’ ACHs being 30 to 60kb from each other. RNA Pol II (RNAPII) ChIA-PET data suggests there is an inclusive hierarchical regulatory structure within the TAD, with substantial interaction between the ACHs of respective pc-genes that is likely facilitated by the 3D conformation. Citation Format: Adam N. Freeman, Michael A. Henderson, Miroslaw K. Kapuscinski, John Hopper, T John Martin. The 3D chromatin structure of the PTHLH region comprises a dynamic hierarchical looping complex that approximates the protein-coding genes and facilitates promoter and enhancer promiscuity. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1975.

Use of a Novel Nonparametric Version of DEPTH to Identify Genomic Regions Associated with Prostate Cancer Risk.

Background: We have developed a genome-wide association study analysis method called DEPTH (DEPendency of association on the number of Top Hits) to identify genomic regions potentially associated with disease by considering overlapping groups of contiguous markers (e.g., SNPs) across the genome. DEPTH is a machine learning algorithm for feature ranking of ultra-high dimensional datasets, built from well-established statistical tools such as bootstrapping, penalized regression, and decision trees. Unlike marginal regression, which considers each SNP individually, the key idea behind DEPTH is to rank groups of SNPs in terms of their joint strength of association with the outcome. Our aim was to compare the performance of DEPTH with that of standard logistic regression analysis. Methods: We selected 1,854 prostate cancer cases and 1,894 controls from the UK for whom 541,129 SNPs were measured using the Illumina Infinium HumanHap550 array. Confirmation was sought using 4,152 cases and 2,874 controls, ascertained from the UK and Australia, for whom 211,155 SNPs were measured using the iCOGS Illumina Infinium array. Results: From the DEPTH analysis, we identified 14 regions associated with prostate cancer risk that had been reported previously, five of which would not have been identified by conventional logistic regression. We also identified 112 novel putative susceptibility regions. Conclusions: DEPTH can reveal new risk-associated regions that would not have been identified using a conventional logistic regression analysis of individual SNPs. Impact: This study demonstrates that the DEPTH algorithm could identify additional genetic susceptibility regions that merit further investigation. Cancer Epidemiol Biomarkers Prev; 25(12); 1619–24. ©2016 AACR.

Abstract 1098: Epistasis analysis of the PTHLH region in European subjects of the iCOGS breast cancer GWAS suggest multiple genes may be implicated in its role in breast cancer

Parathyroid hormone-related protein (PTHrP), the product of the PTHLH gene, has long been implicated in breast cancer. Its expression is thought to favour and, potentially, facilitate metastasis to bone. Paradoxically, a prospective clinical study clinical suggests that its production in primary breast cancers is actually protective, affording a better prognosis than its absence. Multiple recent Genome-Wide Association Studies (GWAS) have confirmed a single susceptibility locus immediately upstream of the PTHLH gene to be associated with breast cancer. This was again reproduced in the recent iCOGS GWAS, which involved over 90,000 European subjects. In spite of a single SNP, rs10771399, being uniquely reported and reproduced by prior studies, comprehensive analysis of the region reveals multiple significant SNPs stretching for hundreds of kbp across the locus. Epistasis analysis of this region versus the entire iCOGS array was performed using Plink!. These results suggest multiple genes may be implicated in the causation of this signal. Many known functional partners of PTHLH were identified, including IPO8, RUNX1, RUNX3, S100P, SOX9, SOX11, and SOX14. Multiple putative functional partners of PTHLH were likewise identified, including PBXIP1, PMVK, KCNN3, STX12, PPP1R8, CAMTA1, PRKCZ, TGIF1, SPHKAP, MYC, and STAT3. When taken together, these results offer novel potential insights into the function of the locus at PTHLH in breast cancer that may be further investigated. Citation Format: Adam N. Freeman, Michael A. Henderson, T John Martin, Enes Makalic, Miroslaw K. Kapuscinski, Daniel Schmidt, John Hopper. Epistasis analysis of the PTHLH region in European subjects of the iCOGS breast cancer GWAS suggest multiple genes may be implicated in its role in breast cancer. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 1098. doi:10.1158/1538-7445.AM2015-1098

Abstract A1-10: The iCOGS breast cancer GWAS reveals 4 unique signals in the PTHLH region in patients of European origin

Parathyroid hormone-related protein (PTHrP), the product of the PTHLH gene, has long been implicated in breast cancer. Its expression is thought to favor and, potentially, facilitate metastasis to bone. Paradoxically, a prospective clinical study clinical suggests that its production in primary breast cancers is actually protective, affording a better prognosis than its absence. Multiple recent Genome-Wide Association Studies (GWAS) have confirmed a single susceptibility locus immediately upstream of the PTHLH gene to be associated with breast cancer. This was again reproduced in the recent iCOGS GWAS, which involved over 90,000 European subjects. In spite of a single SNP, rs10771399, being uniquely reported and reproduced by prior studies, we demonstrate 4 discrete overlapping signals by utilizing forward selection logistic regression and LASSO techniques. Two of these signals are centered and superimposed around rs10771399, ~40kbp upstream of PTHLH, a third lies a further 100kbp upstream, and the fourth lies a further 250kbp upstream over the next gene, CCDC91. While the causation of this signal remains elusive, multiple putatively contributory variations are captured by these signals. Citation Format: Adam N. Freeman, T John Martin, Michael A. Henderson, Enes Makalic, Miroslaw K. Kapuscinski, Daniel F. Schmidt, John Hopper. The iCOGS breast cancer GWAS reveals 4 unique signals in the PTHLH region in patients of European origin. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A1-10.