Search

SummaryPreparing functional genomic (FG) data with diverse assay types and file formats for integration into analysis workflows that interpret genome-wide association and other studies is a significant and time-consuming challenge. Here we introduce hipFG, an automatically customized pipeline for efficient and scalable normalization of heterogenous FG data collections into standardized, indexed, rapidly searchable analysis-ready datasets while accounting for FG datatypes (e.g., chromatin interactions, genomic intervals, quantitative trait loci).Availability and ImplementationhipFG is freely available athttps://bitbucket.org/wanglab-upenn/hipFG. Docker container is available athttps://hub.docker.com/r/wanglab/hipfg.Contactlswang@pennmedicine.upenn.eduSupplementary informationSupplementary data are available as BioRxiv supplemental files.

BackgroundReliable detection and accurate genotyping of structural variants (SVs) and insertion/deletions (indels) from whole-genome sequence (WGS) data is a significant challenge. We present a protocol for variant calling, quality control, call merging, sensitivity analysis, in silico genotyping, and laboratory validation protocols for generating a high-quality deletion call set from whole genome sequences as part of the Alzheimer’s Disease Sequencing Project (ADSP). This dataset contains 578 individuals from 111 families.MethodsWe applied two complementary pipelines (Scalpel and Parliament) for SV/indel calling, break-point refinement, genotyping, and local reassembly to produce a high-quality annotated call set. Sensitivity was measured in sample replicates (N=9) for all callers using in silico variant spike-in for a wide range of event sizes. We focused on deletions because these events were more reliably called. To evaluate caller specificity, we developed a novel metric called the D-score that leverages deletion sharing frequencies within and outside of families to rank recurring deletions. Assessment of overall quality across size bins was measured with the kinship coefficient. Individual callers were evaluated for computational cost, performance, sensitivity, and specificity. Quality of calls were evaluated by Sanger sequencing of predicted loss-of-function (LOF) variants, variants near AD candidate genes, and randomly selected genome-wide deletions ranging from 2 to 17,000 bp.ResultsWe generated a high-quality deletion call set across a wide range of event sizes consisting of 152,301 deletions with an average of 263 per genome. A total of 114 of 146 predicted deletions (78.1%) were validated by Sanger sequencing. Scalpel was more accurate in calling deletions ≤100 bp, whereas for Parliament, sensitivity was improved for deletions > 900 bp. We validated 83.0% (88/106) and 72.5% (37/51) of calls made by Scalpel and Parliament, respectively. Eleven deletions called by both Parliament and Scalpel in the 101-900 bin were tested and all were confirmed by Sanger sequencing.ConclusionsWe developed a flexible protocol to assess the quality of deletion detection across a wide range of sizes. We also generated a truth set of Sanger sequencing validated deletions with precise breakpoints covering a wide spectrum of sizes between 1 and 17,000 bp.

Over 90% of variants are rare, and 50% of them are singletons in the Alzheimer’s Disease Sequencing Project Whole Exome Sequencing (ADSP WES) data. However, either single variant tests or unit-based tests are limited in the statistical power to detect the association between rare variants and phenotypes. To best utilize rare variants and investigate their biological effect, we exam their association with phenotypes in the context of protein. We developed a protein structure-based approach, POKEMON (Protein Optimized Kernel Evaluation of Missense Nucleotides), which evaluates rare missense variants based on their spatial distribution on the protein rather than allele frequency. The hypothesis behind this is that the three-dimensional spatial distribution of variants within a protein structure provides functional context and improves the power of association tests. POKEMON identified four candidate genes from the ADSP WES data, namely two known Alzheimer’s disease (AD) genes (TREM2 and SORL) and two novel genes (DUSP18 and CSF1R). For known AD genes, the signal from the spatial cluster is stable even if we exclude known AD risk variants, indicating the presence of additional low frequency risk variants within these genes. DUSP18 has a cluster of variants primarily shared by case subjects around the ligand-binding domain, and this cluster is further validated in a replication dataset with a larger sample size. POKEMON is an open-source tool available at https://github.com/bushlab-genomics/POKEMON.

Risk for late-onset Alzheimer’s disease (LOAD) is driven by multiple loci primarily identified by genome-wide association studies, many of which are common variants with minor allele frequencies (MAF)> 0.01. To identify additional common and rare LOAD risk variants, we performed a GWAS on 25,170 LOAD subjects and 41,052 cognitively normal controls in 44 datasets from the International Genomics of Alzheimer’s Project (IGAP). Existing genotype data was imputed using the dense, high-resolution Haplotype Reference Consortium (HRC) r1.1 reference panel. Stage 1 associations of P−5 were meta-analyzed with the European Alzheimer’s Disease Biobank (EADB) (n=20,301 cases; 21,839 controls) (stage 2 combined IGAP and EADB). An expanded meta-analysis was performed using a GWAS of parental AD/dementia history in the UK Biobank (UKBB) (n=35,214 cases; 180,791 controls) (stage 3 combined IGAP, EADB, and UKBB). Common variant (MAF≥0.01) associations were identified for 29 loci in stage 2, including novel genome-wide significant associations at TSPAN14 (P=2.33×10−12), SHARPIN (P=1.56×10−9), and ATF5/SIGLEC11 (P=1.03×10−8), and newly significant associations without using AD proxy cases in MTSS1L/IL34 (P=1.80×10−8), APH1B (P=2.10×10−13), and CLNK (P=2.24×10−10). Rare variant (MAFAPOE and TREM2, and a novel association of a rare variant (rs143080277; MAF=0.0054; P=2.69×10−9) in NCK2, further strengthened with the inclusion of UKBB data in stage 3 (P=7.17×10−13). Single-nucleus sequence data shows that NCK2 is highly expressed in amyloid-responsive microglial cells, suggesting a role in LOAD pathology.

MotivationQuerying massive collections of functional genomic and annotation data, linking and summarizing the query results across data sources and data types are important steps in high-throughput genomic and genetic analytical workflows. However, accomplishing these steps is difficult because of the heterogeneity and breadth of data sources, experimental assays, biological conditions (e.g., tissues, cell types), data types, and file formats.ResultsFunctIonaL gEnomics Repository (FILER) is a large-scale, harmonized functional genomics data catalog uniquely providing: 1) streamlined access to >50,000 harmonized, annotated functional genomic and annotation datasets across >20 integrated data sources, >1,100 biological conditions/tissues/cell types, and >20 experimental assays; 2) a scalable, indexing-based genomic querying interface; 3) ability for users to analyze and annotate their own experimental data against reference datasets. This rich resource spans >17 Billion genomic records for both GRCh37/hg19 and GRCh38/hg38 genome builds. FILER scales well with the experimental (query) data size and the number of reference datasets and data sources. When evaluated on large-scale analysis tasks, FILER demonstrated great efficiency as the observed running time for querying 1000x more genomic intervals (106 vs. 103) against all 7×109 hg19 FILER records increased sub-linearly by only a factor of 15x. Together, these features facilitate reproducible research and streamline querying, integrating, and utilizing large-scale functional genomics and annotation data.Availability and implementationFILER can be 1) freely accessed at https://lisanwanglab.org/FILER, 2) deployed on cloud or local servers (https://bitbucket.org/wanglab-upenn/FILER), and 3) integrated with other pipelines using provided scripts.Contactlswang@pennmedicine.upenn.edu

BackgroundRecent reports suggest that the rare apolipoprotein E (APOE) Christchurch mutation and ε2 allele protect against Alzheimer’s disease (AD) pathology by reducing the burden of tau pathology. However, the mechanism(s) underlying the ε2 protective effect linking to tau is largely unknown.MethodsThe role of the ε2 allele in Alzheimer’s disease (AD) was investigated a genome-wide association study (GWAS) for AD among 2,120 ε2 carriers from the Alzheimer Disease Genetics Consortium (ADGC), and then prioritized by gene network analysis, differential gene expression analysis at tissue- and cell-levels as well as methylation profiling of CpG sites, in prefrontal cortex tissue from 761 brains of the Religious Orders Study and Memory and Aging Project (ROSMAP) and the Framingham Heart Study (FHS), Boston University Alzheimer’s Disease Center (BUADC). The levels of two catalytic subunit proteins from protein phosphatase 2A (PPP2CA and PPP2CB) were validated in prefrontal cortex area of 193 of the FHS/BUADC brains. The findings from human autopsied brains were further validated by a co-culture experiment of human isogenic APOE induced pluripotent stem cell (iPSC) derived neurons and astrocytes.ResultsOf the significantly associated loci with AD among APOE ε2 carriers (P−6), PPP2CB (P=1.1×10−7) was the key node in the APOE ε2-related gene network and contained the most significant CpG site (P=7.3×10−4) located 2,814 base pair upstream of the top-ranked GWAS variant. Among APOE ε3/ε4 subjects, the level of Aβ42 was negatively correlated with protein levels of PPP2CA (P=9.9×10−3) and PPP2CB (P=2.4×10−3), and PPP2CA level was correlated with the level of pTau231 level (P=5.3×10−3). Significant correlations were also observed for PPP2CB with complement 4B (C4B) protein levels (P=3.3×10−7) and PPP2CA with cross reactive protein (CRP) levels (P=6.4×10−4). C1q level was not associated with Aβ42, pTau231, PPP2CB, or C4B levels. We confirmed the significant correlation of PPP2CB expression with pTau231/tTau ratio (P=0.01) and C4A/B (P=2.0×10−4) expression observed in brain tissue in a co-culture experiment of iPSC derived neurons and astrocytes.ConclusionWe demonstrated for the first time a molecular link between a tau phosphatase and the classical complement pathway, especially C4, and AD-related tau pathology.

Alzheimer’s Disease (AD) genetics has made substantial progress through genome-wide association studies (GWASs). An up-to-date resource providing harmonized, searchable information on AD genetic variants with linking to genes and supporting functional evidence is needed.We developed the Alzheimer’s Disease Variant Portal (ADVP), an extensive collection of associations curated from >200 GWAS publications from Alzheimer’s Disease Genetics Consortium (ADGC) and other researchers. Publications are reviewed systematically to extract top associations for harmonization and genomic annotation.ADVP V1.0 catalogs 6,990 associations with disease-risk, expression quantitative traits, endophenotypes and neuropathology across >900 loci, >1,800 variants, >80 cohorts, and 8 populations. ADVP integrates with NIAGADS Alzheimer’s GenomicsDB where investigators can cross-reference other functional evidence.ADVP is a valuable resource for investigators to quickly and systematically explore high-confidence AD genetic findings and provides insights into population- and tissue-specific AD genetic architecture. ADVP is continually maintained and enhanced by NIAGADS and is freely accessible (https://advp.niagads.org).

INTRODUCTION: The NIAGADS Alzheimer9s Genomics Database is an interactive knowledgebase for AD genetics that provides access to GWAS summary statistics datasets deposited at NIAGADS, a national genetics data repository for AD and related dementia (ADRD). METHODS: The website makes available >70 genome-wide summary statistics datasets from GWAS and genome sequencing analysis for AD/ADRD. Variants identified from these datasets are mapped to up-to-date variant and gene annotations from a variety of resources and linked to functional genomics data. The database is powered by a big-data optimized relational database and uses ontologies to consistently annotate study designs and phenotypes, facilitating data harmonization and efficient real-time data analysis and variant or gene report generation. RESULTS: Detailed variant reports provide tabular and interactive graphical summaries of known ADRD associations, as well as highlight variants flagged by the Alzheimer9s Disease Sequencing Project (ADSP). Gene reports provide summaries of co-located ADRD risk-associated variants and have been expanded to include meta-analysis results from aggregate association tests performed by the ADSP allowing us to flag genes with genetic-evidence for AD. DISCUSSION: The GenomicsDB makes available >100 million variant annotations, including ~30 million (5 million novel) variants identified as AD-relevant by ADSP, for browsing and real-time mining via the website or programmatically through a REST API. With a newly redesigned, efficient, search interface and comprehensive record pages linking summary statistics to variant and gene annotations, this resource makes these data both accessible and interpretable, establishing itself as valuable tool for AD research.

Approximately 30% of older adults exhibit the neuropathologic features of Alzheimer’s disease (AD) without signs of cognitive impairment. Yet, little is known about the genetic factors that allow these potentially resilient individuals to remain cognitively normal in the face of substantial neuropathology. We performed a large, genome-wide association study (GWAS) of two previously validated metrics of cognitive resilience quantified using a latent variable modeling approach and representing better-than-predicted cognitive performance for a given level of neuropathology. Data were harmonized across 5,108 participants from a clinical trial of AD and three longitudinal cohort studies of cognitive aging. All analyses were run across all participants and repeated restricting the sample to individuals with normal cognition to identify variants at the earliest stages of disease. As expected, all resilience metrics were genetically correlated with cognitive performance and education attainment traits (p-values−20), and we observed novel correlations with neuropsychiatric conditions (p-values−4). Notably, neither resilience metric was genetically correlated with clinical AD (p-values>0.42) nor associated withAPOE(p-values>0.13). In single variant analyses, we observed a genome-wide significant locus among participants with normal cognition on chromosome 18 upstream ofATP8B1(index SNP rs2571244, MAF=0.08, p=2.3×10−8). The top variant at this locus (rs2571244) was significantly associated with methylation in prefrontal cortex tissue at multiple CpG sites, including one just upstream ofATPB81(cg19596477; p=2×10−13). Overall, this comprehensive genetic analysis of resilience implicates a putative role of vascular risk, metabolism, and mental health in protection from the cognitive consequences of neuropathology, while also providing evidence for a novel resilience gene along the bile acid metabolism pathway.Furthermore, the genetic architecture of resilience appears to be distinct from that of clinical AD, suggesting that a shift in focus to molecular contributors to resilience may identify novel pathways for therapeutic targets.

Summary We report Spark-based INFERence of the molecular mechanisms of NOn-coding genetic variants (SparkINFERNO), a scalable bioinformatics pipeline characterizing non-coding genome-wide association study (GWAS) association findings. SparkINFERNO prioritizes causal variants underlying GWAS association signals and reports relevant regulatory elements, tissue contexts and plausible target genes they affect. To achieve this, the SparkINFERNO algorithm integrates GWAS summary statistics with large-scale collection of functional genomics datasets spanning enhancer activity, transcription factor binding, expression quantitative trait loci and other functional datasets across more than 400 tissues and cell types. Scalability is achieved by an underlying API implemented using Apache Spark and Giggle-based genomic indexing. We evaluated SparkINFERNO on large GWASs and show that SparkINFERNO is more than 60 times efficient and scales with data size and amount of computational resources. Availability and implementation SparkINFERNO runs on clusters or a single server with Apache Spark environment, and is available at https://bitbucket.org/wanglab-upenn/SparkINFERNO or https://hub.docker.com/r/wanglab/spark-inferno. Contact lswang@pennmedicine.upenn.edu Supplementary information Supplementary data are available at Bioinformatics online

Each additional copy of the apolipoprotein E4 (APOE4) allele is associated with a higher risk of Alzheimer’s dementia, such that APOE4 homozygotes have a particularly high risk. While the APOE2 allele is associated with a lower risk of Alzheimer’s dementia, it is not yet known whether APOE2 homozygotes have a particularly low risk. We generated Alzheimer’s dementia odds ratios and other findings in more than 5,000 clinically characterized and neuropathologically characterized Alzheimer’s dementia cases and controls. APOE2/2 was associated with exceptionally low Alzheimer’s dementia odds ratios compared to APOE2/3, 3/3 and 4/4, and the impact of APOE2 and APOE4 gene dose was significantly greater in the neuropathologically confirmed group than in more than 24,000 neuropathologically unconfirmed cases and controls. Finding and targeting the factors by which APOE and its variants influence Alzheimer’s disease could have a major impact on the understanding, treatment and prevention of this terrible disease.

Background Potentially modifiable risk factors have been associated with Alzheimer’s disease (AD). However, the causality of these risk factors on AD is unclear. Using Mendelian randomization we evaluated the causal effect of potentially modifiable risk factors on AD and its associated endophenotypes to inform the development of lifestyle interventions that could reduce risk of developing AD. Methods and Findings Genetic instruments for the exposures were selected from genome wide association studies (GWAS) for traits previously linked to AD in observational studies including alcohol intake, physical activity, lipid traits, blood pressure traits, type 2 diabetes, body-mass index (BMI), depression, sleep, social isolation, smoking, oily fish intake, and educational attainment. The outcomes included AD status, AD age of onset survival (AAOS), hippocampal volume, CSF levels of Aβ 42 , tau, and ptau 181 , and neuropathological burden of neuritic plaques, neurofibrillary tangles, and vascular brain injury (VBI). MR estimates were calculated using an inverse variance weighted approach. Genetically predicted educational attainment (OR [CI]: 0.7 [0.63, 0.78]), diastolic blood pressure (DBP) (OR [CI]: 0.99 [0.98, 0.99]), systolic blood pressure (SBP) (OR [CI]: 0.99 [0.99, 1]) and physical activity (OR [CI]: 2.5 [1.47, 4.23]) were causally associated with AD risk. Genetically predicted BMI (HR [CI]: 1.13 [1.07, 1.2]) and educational attainment (HR [CI]: 0.74 [0.68, 0.82]) were causally associated with AAOS. Genetically predicted alcohol consumption (β [CI]: −0.15 [−0.25, −0.05]), broad depressive symptoms (β [CI]: 0.5 [0.2, 0.8]), major depressive disorder (β [CI]: 0.12 [0.05, 0.18]) and physical activity were causally associated with CSF Aβ 42 (β [CI]: 0.25 [0.1, 0.39]). Genetically predicted DBP was causally associated with CSF total Tau (β [CI]: −0.005 [−0.007, −0.002]). Increased risk of VBI were observed for genetically predicted DBP (OR [CI]: 1.05 [1.02, 1.08]), SBP (OR [CI]: 1.06 [1.03, 1.1]), and pulse pressure (OR [CI]: 1.03 [1.01, 1.05]). Increased risk of neuritic plaque burden were observed for genetically predicted LDL-cholesterol (OR [CI]: 1.87 [1.3, 2.69]) and total cholesterol (OR [CI]: 2.03 [1.44, 2.85]). Genetically predicted insomnia symptoms (β [CI]: −0.2 [−0.34, −0.06]) and total cholesterol were associated (β [CI]: −0.06 [−0.1, −0.03]) with hippocampal volume. Potential limitations include weak instrument bias, non-homogenous samples and other implicit limitations of MR analysis. Conclusions Demonstration of a causal relationship between blood pressure, cholesterol levels, BMI, depression, insomnia symptoms, physical activity and educational attainment on the AD phenome strongly support public health programs to educate the public about these preventable causes of AD.

Most regulatory chromatin interactions are mediated by various transcription factors (TFs) and involve physically-interacting elements such as enhancers, insulators, or promoters. To map these elements and interactions, we developed HIPPIE2 which analyzes raw reads from high-throughput chromosome conformation (Hi-C) experiments to identify fine-scale physically-interacting regions (PIRs). Unlike standard genome binning approaches (e.g., 10K-1Mbp bins), HIPPIE2 dynamically calls physical locations of PIRs with better precision and higher resolution based on the pattern of restriction events and relative locations of interacting sites inferred from the sequencing readout.We applied HIPPIE2 to in situ Hi-C datasets across 6 human cell lines (GM12878, IMR90, K562, HMEC, HUVEC, NHEK) with matched ENCODE and Roadmap functional genomic data. HIPPIE2 detected 1,042,738 distinct PIRs across cell lines, with high resolution (average PIR length of 1,006bps) and high reproducibility (92.3% in GM12878 replicates). 32.8% of PIRs were shared among cell lines. PIRs are enriched for epigenetic marks (H3K27ac, H3K4me1) and open chromatin, suggesting active regulatory roles. HIPPIE2 identified 2.8M significant intrachromosomal PIR–PIR interactions, 27.2% of which were enriched for TF binding sites. 50,608 interactions were enhancer–promoter interactions and were enriched for 33 TFs (31 in enhancers/29 in promoters), several of which are known to mediate DNA looping/long-distance regulation. 29 TFs were enriched in >1 cell line and 4 were cell line-specific. These findings demonstrate that the dynamic approach used in HIPPIE2 (https://bitbucket.com/wanglab-upenn/HIPPIE2) characterizes PIR–PIR interactions with high resolution and reproducibility.

The Alzheimer’s Disease Sequencing Project (ADSP) performed whole genome sequencing (WGS) of 584 subjects from 111 multiplex families at three sequencing centers. Genotype calling of single nucleotide variants (SNVs) and insertion-deletion variants (indels) was performed centrally using GATK-HaplotypeCaller and Atlas V2. The ADSP Quality Control (QC) Working Group applied QC protocols to project-level variant call format files (VCFs) from each pipeline, and developed and implemented a novel protocol, termed “consensus calling,” to combine genotype calls from both pipelines into a single high-quality set. QC was applied to autosomal bi-allelic SNVs and indels, and included pipeline-recommended QC filters, variant-level QC, and sample-level QC. Low-quality variants or genotypes were excluded, and sample outliers were noted. Quality was assessed by examining Mendelian inconsistencies (MIs) among 67 parent-offspring pairs, and MIs were used to establish additional genotype-specific filters for GATK calls. After QC, 578 subjects remained. Pipeline-specific QC excluded ~12.0% of GATK and 14.5% of Atlas SNVs. Between pipelines, ~91% of SNV genotypes across all QCed variants were concordant; 4.23% and 4.56% of genotypes were exclusive to Atlas or GATK, respectively; the remaining ~0.01% of discordant genotypes were excluded. For indels, variant-level QC excluded ~36.8% of GATK and 35.3% of Atlas indels. Between pipelines, ~55.6% of indel genotypes were concordant; while 10.3% and 28.3% were exclusive to Atlas or GATK, respectively; and ~0.29% of discordant genotypes were. The final WGS consensus dataset contains 27,896,774 SNVs and 3,133,926 indels and is publicly available.AbbreviationsAD, Alzheimer’s disease; QC, Quality Control; LSSAC, Large-Scale Sequencing and Analysis Center; Broad, Broad Institute Genomics Service; Baylor, Baylor College of Medicine Human Genome Sequencing Center; WashU, Washington University-St. Louis McDonnell Genome Institute; WGS, whole genome sequencing; WES, whole exome sequencing; indel, insertion-deletion variants; VCF, variant control format; MI, Mendelian inconsistency; MC, Mendelian consistency; GWAS, genome-wide association study; VR, referent allele read depth; DP, overall read depth; MS, mapping score; GQ, genotype quality score; Ti/Tv, Transition/Transversion; CS, concordance code

MotivationCopy number variations (CNVs) are gains and losses of DNA segments and have been associated with disease. Many large-scale genetic association studies are performing CNV analysis using whole exome sequencing (WES) and whole genome sequencing (WGS). In many of these studies, previous SNP-array data are available. An integrated cross-platform analysis is expected to improve resolution and accuracy, yet there is no tool for effectively combining data from sequencing and array platforms. The detection of CNVs using sequencing data alone can also be further improved by the utilization of allele-specific reads.ResultsWe propose a statistical framework, integrated Copy Number Variation detection algorithm (iCNV), which can be applied to multiple study designs: WES only, WGS only, SNP array only, or any combination of SNP and sequencing data. iCNV applies platform specific normalization, utilizes allele specific reads from sequencing and integrates matched NGS and SNP-array data by a Hidden Markov Model (HMM). We compare integrated two-platform CNV detection using iCNV to naive intersection or union of platforms and show that iCNV increases sensitivity and robustness. We also assess the accuracy of iCNV on WGS data only, and show that the utilization of allele-specific reads improve CNV detection accuracy compared to existing methods.Availabilityhttps://github.com/zhouzilu/iCNVContactnzh@wharton.upenn.edu, zhouzilu@mail.med.upenn.eduSupplementary informationSupplementary data are available at Bioinformatics online.

Cellular senescence is accompanied by dramatic changes in chromatin structure and gene expression. Using Saccharomyces cerevisiae mutants lacking telomerase (tlc1Δ) to model senescence, we found that with critical telomere shortening, the telomere-binding protein Rap1 (repressor activator protein 1) relocalizes to the upstream promoter regions of hundreds of new target genes. The set of new Rap1 targets at senescence (NRTS) is preferentially activated at senescence, and experimental manipulations of Rap1 levels indicate that it contributes directly to NRTS activation. A notable subset of NRTS includes the core histone-encoding genes; we found that Rap1 contributes to their repression and that histone protein levels decline at senescence. Rap1 and histones also display a target site-specific antagonism that leads to diminished nucleosome occupancy at the promoters of up-regulated NRTS. This antagonism apparently impacts the rate of senescence because underexpression of Rap1 or overexpression of the core histones delays senescence. Rap1 relocalization is not a simple consequence of lost telomere-binding sites, but rather depends on the Mec1 checkpoint kinase. Rap1 relocalization is thus a novel mechanism connecting DNA damage responses (DDRs) at telomeres to global changes in chromatin and gene expression while driving the pace of senescence.