Comprehensive Characterization of Cancer Driver Genes and Mutations.

Identifying molecular cancer drivers is critical for precision oncology. Multiple advanced algorithms to identify drivers now exist, but systematic attempts to combine and optimize them on large datasets are few. We report a PanCancer and PanSoftware analysis spanning 9,423 tumor exomes (comprising all 33 of The Cancer Genome Atlas projects) and using 26 computational tools to catalog driver genes and mutations. We identify 299 driver genes with implications regarding their anatomical sites and cancer/cell types. Sequence- and structure-based analyses identified >3,400 putative missense driver mutations supported by multiple lines of evidence. Experimental validation confirmed 60%-85% of predicted mutations as likely drivers. We found that >300 MSI tumors are associated with high PD-1/PD-L1, and 57% of tumors analyzed harbor putative clinically actionable events. Our study represents the most comprehensive discovery of cancer genes and mutations to date and will serve as a blueprint for future biological and clinical endeavors.

3D genomic mapping reveals multifocality of human pancreatic precancers.

Pancreatic intraepithelial neoplasias (PanINs) are the most common precursors of pancreatic cancer, but their small size and inaccessibility in humans make them challenging to study1. Critically, the number, dimensions and connectivity of human PanINs remain largely unknown, precluding important insights into early cancer development. Here, we provide a microanatomical survey of human PanINs by analysing 46 large samples of grossly normal human pancreas with a machine-learning pipeline for quantitative 3D histological reconstruction at single-cell resolution. To elucidate genetic relationships between and within PanINs, we developed a workflow in which 3D modelling guides multi-region microdissection and targeted and whole-exome sequencing. From these samples, we calculated a mean burden of 13 PanINs per cm3 and extrapolated that the normal intact adult pancreas harbours hundreds of PanINs, almost all with oncogenic KRAS hotspot mutations. We found that most PanINs originate as independent clones with distinct somatic mutation profiles. Some spatially continuous PanINs were found to contain multiple KRAS mutations; computational and in situ analyses demonstrated that different KRAS mutations localize to distinct cell subpopulations within these neoplasms, indicating their polyclonal origins. The extensive multifocality and genetic heterogeneity of PanINs raises important questions about mechanisms that drive precancer initiation and confer differential progression risk in the human pancreas. This detailed 3D genomic mapping of molecular alterations in human PanINs provides an empirical foundation for early detection and rational interception of pancreatic cancer.

Calibration of computational tools for missense variant pathogenicity classification and ClinGen recommendations for PP3/BP4 criteria.

Recommendations from the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) for interpreting sequence variants specify the use of computational predictors as "supporting" level of evidence for pathogenicity or benignity using criteria PP3 and BP4, respectively. However, score intervals defined by tool developers, and ACMG/AMP recommendations that require the consensus of multiple predictors, lack quantitative support. Previously, we described a probabilistic framework that quantified the strengths of evidence (supporting, moderate, strong, very strong) within ACMG/AMP recommendations. We have extended this framework to computational predictors and introduce a new standard that converts a tool's scores to PP3 and BP4 evidence strengths. Our approach is based on estimating the local positive predictive value and can calibrate any computational tool or other continuous-scale evidence on any variant type. We estimate thresholds (score intervals) corresponding to each strength of evidence for pathogenicity and benignity for thirteen missense variant interpretation tools, using carefully assembled independent data sets. Most tools achieved supporting evidence level for both pathogenic and benign classification using newly established thresholds. Multiple tools reached score thresholds justifying moderate and several reached strong evidence levels. One tool reached very strong evidence level for benign classification on some variants. Based on these findings, we provide recommendations for evidence-based revisions of the PP3 and BP4 ACMG/AMP criteria using individual tools and future assessment of computational methods for clinical interpretation.

Estimation of cancer cell fractions and clone trees from multi-region sequencing of tumors.

MOTIVATION: Multi-region sequencing of solid tumors can improve our understanding of intratumor subclonal diversity and the evolutionary history of mutational events. Due to uncertainty in clonal composition and the multitude of possible ancestral relationships between clones, elucidating the most probable relationships from bulk tumor sequencing poses statistical and computational challenges. RESULTS: We developed a Bayesian hierarchical model called PICTograph to model uncertainty in assigning mutations to subclones, to enable posterior distributions of cancer cell fractions (CCFs) and to visualize the most probable ancestral relationships between subclones. Compared with available methods, PICTograph provided more consistent and accurate estimates of CCFs and improved tree inference over a range of simulated clonal diversity. Application of PICTograph to multi-region whole-exome sequencing of tumors from individuals with pancreatic cancer precursor lesions confirmed known early-occurring mutations and indicated substantial molecular diversity, including 6-12 distinct subclones and intra-sample mixing of subclones. Using ensemble-based visualizations, we highlight highly probable evolutionary relationships recovered in multiple models. PICTograph provides a useful approximation to evolutionary inference from cross-sectional multi-region sequencing, particularly for complex cases. AVAILABILITY AND IMPLEMENTATION: https://github.com/KarchinLab/pictograph. The data underlying this article will be shared on reasonable request to the corresponding author. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Assessing aneuploidy with repetitive element sequencing.

We report a sensitive PCR-based assay called Repetitive Element AneupLoidy Sequencing System (RealSeqS) that can detect aneuploidy in samples containing as little as 3 pg of DNA. Using a single primer pair, we amplified ∼350,000 amplicons distributed throughout the genome. Aneuploidy was detected in 49% of liquid biopsies from a total of 883 nonmetastatic, clinically detected cancers of the colorectum, esophagus, liver, lung, ovary, pancreas, breast, or stomach. Combining aneuploidy with somatic mutation detection and eight standard protein biomarkers yielded a median sensitivity of 80% in these eight cancer types, while only 1% of 812 healthy controls scored positive.

Multiregion whole-exome sequencing of intraductal papillary mucinous neoplasms reveals frequent somatic KLF4 mutations predominantly in low-grade regions.

OBJECTIVE: Intraductal papillary mucinous neoplasms (IPMNs) are non-invasive precursor lesions that can progress to invasive pancreatic cancer and are classified as low-grade or high-grade based on the morphology of the neoplastic epithelium. We aimed to compare genetic alterations in low-grade and high-grade regions of the same IPMN in order to identify molecular alterations underlying neoplastic progression. DESIGN: We performed multiregion whole exome sequencing on tissue samples from 17 IPMNs with both low-grade and high-grade dysplasia (76 IPMN regions, including 49 from low-grade dysplasia and 27 from high-grade dysplasia). We reconstructed the phylogeny for each case, and we assessed mutations in a novel driver gene in an independent cohort of 63 IPMN cyst fluid samples. RESULTS: Our multiregion whole exome sequencing identified KLF4, a previously unreported genetic driver of IPMN tumorigenesis, with hotspot mutations in one of two codons identified in >50% of the analyzed IPMNs. Mutations in KLF4 were significantly more prevalent in low-grade regions in our sequenced cases. Phylogenetic analyses of whole exome sequencing data demonstrated diverse patterns of IPMN initiation and progression. Hotspot mutations in KLF4 were also identified in an independent cohort of IPMN cyst fluid samples, again with a significantly higher prevalence in low-grade IPMNs. CONCLUSION: Hotspot mutations in KLF4 occur at high prevalence in IPMNs. Unique among pancreatic driver genes, KLF4 mutations are enriched in low-grade IPMNs. These data highlight distinct molecular features of low-grade and high-grade dysplasia and suggest diverse pathways to high-grade dysplasia via the IPMN pathway.

Assessment of the Clinical Relevance of BRCA2 Missense Variants by Functional and Computational Approaches.

Many variants of uncertain significance (VUS) have been identified in BRCA2 through clinical genetic testing. VUS pose a significant clinical challenge because the contribution of these variants to cancer risk has not been determined. We conducted a comprehensive assessment of VUS in the BRCA2 C-terminal DNA binding domain (DBD) by using a validated functional assay of BRCA2 homologous recombination (HR) DNA-repair activity and defined a classifier of variant pathogenicity. Among 139 variants evaluated, 54 had ?99% probability of pathogenicity, and 73 had ?95% probability of neutrality. Functional assay results were compared with predictions of variant pathogenicity from the Align-GVGD protein-sequence-based prediction algorithm, which has been used for variant classification. Relative to the HR assay, Align-GVGD significantly (p < 0.05) over-predicted pathogenic variants. We subsequently combined functional and Align-GVGD prediction results in a Bayesian hierarchical model (VarCall) to estimate the overall probability of pathogenicity for each VUS. In addition, to predict the effects of all other BRCA2 DBD variants and to prioritize variants for functional studies, we used the endoPhenotype-Optimized Sequence Ensemble (ePOSE) algorithm to train classifiers for BRCA2 variants by using data from the HR functional assay. Together, the results show that systematic functional assays in combination with in silico predictors of pathogenicity provide robust tools for clinical annotation of BRCA2 VUS.

The genomic landscape of response to EGFR blockade in colorectal cancer.

Colorectal cancer is the third most common cancer worldwide, with 1.2 million patients diagnosed annually. In late-stage colorectal cancer, the most commonly used targeted therapies are the monoclonal antibodies cetuximab and panitumumab, which prevent epidermal growth factor receptor (EGFR) activation. Recent studies have identified alterations in KRAS and other genes as likely mechanisms of primary and secondary resistance to anti-EGFR antibody therapy. Despite these efforts, additional mechanisms of resistance to EGFR blockade are thought to be present in colorectal cancer and little is known about determinants of sensitivity to this therapy. To examine the effect of somatic genetic changes in colorectal cancer on response to anti-EGFR antibody therapy, here we perform complete exome sequence and copy number analyses of 129 patient-derived tumour grafts and targeted genomic analyses of 55 patient tumours, all of which were KRAS wild-type. We analysed the response of tumours to anti-EGFR antibody blockade in tumour graft models and in clinical settings and functionally linked therapeutic responses to mutational data. In addition to previously identified genes, we detected mutations in ERBB2, EGFR, FGFR1, PDGFRA, and MAP2K1 as potential mechanisms of primary resistance to this therapy. Novel alterations in the ectodomain of EGFR were identified in patients with acquired resistance to EGFR blockade. Amplifications and sequence changes in the tyrosine kinase receptor adaptor gene IRS2 were identified in tumours with increased sensitivity to anti-EGFR therapy. Therapeutic resistance to EGFR blockade could be overcome in tumour graft models through combinatorial therapies targeting actionable genes. These analyses provide a systematic approach to evaluating response to targeted therapies in human cancer, highlight new mechanisms of responsiveness to anti-EGFR therapies, and delineate new avenues for intervention in managing colorectal cancer.

Detection of aneuploidy in patients with cancer through amplification of long interspersed nucleotide elements (LINEs).

Aneuploidy is a feature of most cancer cells, and a myriad of approaches have been developed to detect it in clinical samples. We previously described primers that could be used to amplify ∼38,000 unique long interspersed nucleotide elements (LINEs) from throughout the genome. Here we have developed an approach to evaluate the sequencing data obtained from these amplicons. This approach, called Within-Sample AneupLoidy DetectiOn (WALDO), employs supervised machine learning to detect the small changes in multiple chromosome arms that are often present in cancers. We used WALDO to search for chromosome arm gains and losses in 1,677 tumors and in 1,522 liquid biopsies of blood from cancer patients or normal individuals. Aneuploidy was detected in 95% of cancer biopsies and in 22% of liquid biopsies. Using single-nucleotide polymorphisms within the amplified LINEs, WALDO concomitantly assesses allelic imbalances, microsatellite instability, and sample identification. WALDO can be used on samples containing only a few nanograms of DNA and as little as 1% neoplastic content and has a variety of applications in cancer diagnostics and forensic science.

Intraductal Papillary Mucinous Neoplasms Arise From Multiple Independent Clones, Each With Distinct Mutations.

BACKGROUND & AIMS: Intraductal papillary mucinous neoplasms (IPMNs) are lesions that can progress to invasive pancreatic cancer and constitute an important system for studies of pancreatic tumorigenesis. We performed comprehensive genomic analyses of entire IPMNs to determine the diversity of somatic mutations in genes that promote tumorigenesis. METHODS: We microdissected neoplastic tissues from 6-24 regions each of 20 resected IPMNs, resulting in 227 neoplastic samples that were analyzed by capture-based targeted sequencing. Somatic mutations in genes associated with pancreatic tumorigenesis were assessed across entire IPMN lesions, and the resulting data were supported by evolutionary modeling, whole-exome sequencing, and in situ detection of mutations. RESULTS: We found a high prevalence of heterogeneity among mutations in IPMNs. Heterogeneity in mutations in KRAS and GNAS was significantly more prevalent in IPMNs with low-grade dysplasia than in IPMNs with high-grade dysplasia (P < .02). Whole-exome sequencing confirmed that IPMNs contained multiple independent clones, each with distinct mutations, as originally indicated by targeted sequencing and evolutionary modeling. We also found evidence for convergent evolution of mutations in RNF43 and TP53, which are acquired during later stages of tumorigenesis. CONCLUSIONS: In an analysis of the heterogeneity of mutations throughout IPMNs, we found that early-stage IPMNs contain multiple independent clones, each with distinct mutations, indicating their polyclonal origin. These findings challenge the model in which pancreatic neoplasms arise from a single clone. Increasing our understanding of the mechanisms of IPMN polyclonality could lead to strategies to identify patients at increased risk for pancreatic cancer.

Single-cell sequencing defines genetic heterogeneity in pancreatic cancer precursor lesions.

Intraductal papillary mucinous neoplasms (IPMNs) are precursors to pancreatic cancer; however, little is known about genetic heterogeneity in these lesions. The objective of this study was to characterize genetic heterogeneity in IPMNs at the single-cell level. We isolated single cells from fresh tissue from ten IPMNs, followed by whole genome amplification and targeted next-generation sequencing of pancreatic driver genes. We then determined single-cell genotypes using a novel multi-sample mutation calling algorithm. Our analyses revealed that different mutations in the same driver gene frequently occur in the same IPMN. Two IPMNs had multiple mutations in the initiating driver gene KRAS that occurred in unique tumor clones, suggesting the possibility of polyclonal origin or an unidentified initiating event preceding this critical mutation. Multiple mutations in later-occurring driver genes were also common and were frequently localized to unique tumor clones, raising the possibility of convergent evolution of these genetic events in pancreatic tumorigenesis. Single-cell sequencing of IPMNs demonstrated genetic heterogeneity with respect to early and late occurring driver gene mutations, suggesting a more complex pattern of tumor evolution than previously appreciated in these lesions. Copyright © 2018 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Assessing computational predictions of the phenotypic effect of cystathionine-beta-synthase variants.

Accurate prediction of the impact of genomic variation on phenotype is a major goal of computational biology and an important contributor to personalized medicine. Computational predictions can lead to a better understanding of the mechanisms underlying genetic diseases, including cancer, but their adoption requires thorough and unbiased assessment. Cystathionine-beta-synthase (CBS) is an enzyme that catalyzes the first step of the transsulfuration pathway, from homocysteine to cystathionine, and in which variations are associated with human hyperhomocysteinemia and homocystinuria. We have created a computational challenge under the CAGI framework to evaluate how well different methods can predict the phenotypic effect(s) of CBS single amino acid substitutions using a blinded experimental data set. CAGI participants were asked to predict yeast growth based on the identity of the mutations. The performance of the methods was evaluated using several metrics. The CBS challenge highlighted the difficulty of predicting the phenotype of an ex vivo system in a model organism when classification models were trained on human disease data. We also discuss the variations in difficulty of prediction for known benign and deleterious variants, as well as identify methodological and experimental constraints with lessons to be learned for future challenges.

Network Analysis of Protein Adaptation: Modeling the Functional Impact of Multiple Mutations.

The evolution of new biochemical activities frequently involves complex dependencies between mutations and rapid evolutionary radiation. Mutation co-occurrence and covariation have previously been used to identify compensating mutations that are the result of physical contacts and preserve protein function and fold. Here, we model pairwise functional dependencies and higher order interactions that enable evolution of new protein functions. We use a network model to find complex dependencies between mutations resulting from evolutionary trade-offs and pleiotropic effects. We present a method to construct these networks and to identify functionally interacting mutations in both extant and reconstructed ancestral sequences (Network Analysis of Protein Adaptation). The time ordering of mutations can be incorporated into the networks through phylogenetic reconstruction. We apply NAPA to three distantly homologous ß-lactamase protein clusters (TEM, CTX-M-3, and OXA-51), each of which has experienced recent evolutionary radiation under substantially different selective pressures. By analyzing the network properties of each protein cluster, we identify key adaptive mutations, positive pairwise interactions, different adaptive solutions to the same selective pressure, and complex evolutionary trajectories likely to increase protein fitness. We also present evidence that incorporating information from phylogenetic reconstruction and ancestral sequence inference can reduce the number of spurious links in the network, whereas preserving overall network community structure. The analysis does not require structural or biochemical data. In contrast to function-preserving mutation dependencies, which are frequently from structural contacts, gain-of-function mutation dependencies are most commonly between residues distal in protein structure.

Evaluating the evaluation of cancer driver genes.

Sequencing has identified millions of somatic mutations in human cancers, but distinguishing cancer driver genes remains a major challenge. Numerous methods have been developed to identify driver genes, but evaluation of the performance of these methods is hindered by the lack of a gold standard, that is, bona fide driver gene mutations. Here, we establish an evaluation framework that can be applied to driver gene prediction methods. We used this framework to compare the performance of eight such methods. One of these methods, described here, incorporated a machine-learning-based ratiometric approach. We show that the driver genes predicted by each of the eight methods vary widely. Moreover, the P values reported by several of the methods were inconsistent with the uniform values expected, thus calling into question the assumptions that were used to generate them. Finally, we evaluated the potential effects of unexplained variability in mutation rates on false-positive driver gene predictions. Our analysis points to the strengths and weaknesses of each of the currently available methods and offers guidance for improving them in the future.

IPMNs with co-occurring invasive cancers: neighbours but not always relatives.

OBJECTIVE: Intraductal papillary mucinous neoplasms (IPMNs) are precursor lesions that can give rise to invasive pancreatic carcinoma. Although approximately 8% of patients with resected pancreatic ductal adenocarcinoma have a co-occurring IPMN, the precise genetic relationship between these two lesions has not been systematically investigated. DESIGN: We analysed all available patients with co-occurring IPMN and invasive intrapancreatic carcinoma over a 10-year period at a single institution. For each patient, we separately isolated DNA from the carcinoma, adjacent IPMN and distant IPMN and performed targeted next generation sequencing of a panel of pancreatic cancer driver genes. We then used the identified mutations to infer the relatedness of the IPMN and co-occurring invasive carcinoma in each patient. RESULTS: We analysed co-occurring IPMN and invasive carcinoma from 61 patients with IPMN/ductal adenocarcinoma as well as 13 patients with IPMN/colloid carcinoma and 7 patients with IPMN/carcinoma of the ampullary region. Of the patients with co-occurring IPMN and ductal adenocarcinoma, 51% were likely related. Surprisingly, 18% of co-occurring IPMN and ductal adenocarcinomas were likely independent, suggesting that the carcinoma arose from an independent precursor. By contrast, all colloid carcinomas were likely related to their associated IPMNs. In addition, these analyses showed striking genetic heterogeneity in IPMNs, even with respect to well-characterised driver genes. CONCLUSION: This study demonstrates a higher prevalence of likely independent co-occurring IPMN and ductal adenocarcinoma than previously appreciated. These findings have important implications for molecular risk stratification of patients with IPMN.

Bridging the Chromosome-centric and Biology/Disease-driven Human Proteome Projects: Accessible and Automated Tools for Interpreting the Biological and Pathological Impact of Protein Sequence Variants Detected via Proteogenomics.

The Chromosome-centric Human Proteome Project (C-HPP) seeks to comprehensively characterize all protein products coded by the genome, including those expressed sequence variants confirmed via proteogenomics methods. The closely related Biology/Disease-driven Human Proteome Project (B/D-HPP) seeks to understand the biological and pathological associations of expressed protein products, especially those carrying sequence variants that may be drivers of disease. To achieve these objectives, informatics tools are required that interpret potential functional or disease implications of variant protein sequence detected via proteogenomics. Toward this end, we have developed an automated workflow within the Galaxy for Proteomics (Galaxy-P) platform, which leverages the Cancer-Related Analysis of Variants Toolkit (CRAVAT) and makes it interoperable with proteogenomic results. Protein sequence variants confirmed by proteogenomics are assessed for potential structure-function effects as well as associations with cancer using CRAVAT's rich suite of functionalities, including visualization of results directly within the Galaxy user interface. We demonstrate the effectiveness of this workflow on proteogenomic results generated from an MCF7 breast cancer cell line. Our free and open software should enable improved interpretation of the functional and pathological effects of protein sequence variants detected via proteogenomics, acting as a bridge between the C-HPP and B/D-HPP.

Matching phenotypes to whole genomes: Lessons learned from four iterations of the personal genome project community challenges.

The advent of next-generation sequencing has dramatically decreased the cost for whole-genome sequencing and increased the viability for its application in research and clinical care. The Personal Genome Project (PGP) provides unrestricted access to genomes of individuals and their associated phenotypes. This resource enabled the Critical Assessment of Genome Interpretation (CAGI) to create a community challenge to assess the bioinformatics community's ability to predict traits from whole genomes. In the CAGI PGP challenge, researchers were asked to predict whether an individual had a particular trait or profile based on their whole genome. Several approaches were used to assess submissions, including ROC AUC (area under receiver operating characteristic curve), probability rankings, the number of correct predictions, and statistical significance simulations. Overall, we found that prediction of individual traits is difficult, relying on a strong knowledge of trait frequency within the general population, whereas matching genomes to trait profiles relies heavily upon a small number of common traits including ancestry, blood type, and eye color. When a rare genetic disorder is present, profiles can be matched when one or more pathogenic variants are identified. Prediction accuracy has improved substantially over the last 6 years due to improved methodology and a better understanding of features.

Missense variants in CFTR nucleotide-binding domains predict quantitative phenotypes associated with cystic fibrosis disease severity.

Predicting the impact of genetic variation on human health remains an important and difficult challenge. Often, algorithmic classifiers are tasked with predicting binary traits (e.g. positive or negative for a disease) from missense variation. Though useful, this arrangement is limiting and contrived, because human diseases often comprise a spectrum of severities, rather than a discrete partitioning of patient populations. Furthermore, labeling variants as causal or benign can be error prone, which is problematic for training supervised learning algorithms (the so-called garbage in, garbage out phenomenon). We explore the potential value of training classifiers using continuous-valued quantitative measurements, rather than binary traits. Using 20 variants from cystic fibrosis transmembrane conductance regulator (CFTR) nucleotide-binding domains and six quantitative measures of cystic fibrosis (CF) severity, we trained classifiers to predict CF severity from CFTR variants. Employing cross validation, classifier prediction and measured clinical/functional values were significantly correlated for four of six quantitative traits (correlation P-values from 1.35 × 10(-4) to 4.15 × 10(-3)). Classifiers were also able to stratify variants by three clinically relevant risk categories with 85-100% accuracy, depending on which of the six quantitative traits was used for training. Finally, we characterized 11 additional CFTR variants using clinical sweat chloride testing, two functional assays, or all three diagnostics, and validated our classifier using blind prediction. Predictions were within the measured sweat chloride range for seven of eight variants, and captured the differential impact of specific variants on the two functional assays. This work demonstrates a promising and novel framework for assessing the impact of genetic variation.

Proteins linked to autosomal dominant and autosomal recessive disorders harbor characteristic rare missense mutation distribution patterns.

The role of rare missense variants in disease causation remains difficult to interpret. We explore whether the clustering pattern of rare missense variants (MAF < 0.01) in a protein is associated with mode of inheritance. Mutations in genes associated with autosomal dominant (AD) conditions are known to result in either loss or gain of function, whereas mutations in genes associated with autosomal recessive (AR) conditions invariably result in loss-of-function. Loss-of-function mutations tend to be distributed uniformly along protein sequence, whereas gain-of-function mutations tend to localize to key regions. It has not previously been ascertained whether these patterns hold in general for rare missense mutations. We consider the extent to which rare missense variants are located within annotated protein domains and whether they form clusters, using a new unbiased method called CLUstering by Mutation Position. These approaches quantified a significant difference in clustering between AD and AR diseases. Proteins linked to AD diseases exhibited more clustering of rare missense mutations than those linked to AR diseases (Wilcoxon P = 5.7 × 10(-4), permutation P = 8.4 × 10(-4)). Rare missense mutation in proteins linked to either AD or AR diseases was more clustered than controls (1000G) (Wilcoxon P = 2.8 × 10(-15) for AD and P = 4.5 × 10(-4) for AR, permutation P = 3.1 × 10(-12) for AD and P = 0.03 for AR). The differences in clustering patterns persisted even after removal of the most prominent genes. Testing for such non-random patterns may reveal novel aspects of disease etiology in large sample studies.