FUN-LDA: A Latent Dirichlet Allocation Model for Predicting Tissue-Specific Functional Effects of Noncoding Variation: Methods and Applications

Polygenic scores and their applications in kidney disease

Genome-wide association studies (GWAS) have uncovered thousands of risk variants that individually have small effects on the risk of human diseases, including chronic kidney disease, type 2 diabetes, heart diseases and inflammatory disorders, but cumulatively explain a substantial fraction of disease risk, underscoring the complexity and pervasive polygenicity of common disorders. This complexity poses unique challenges to the clinical translation of GWAS findings. Polygenic scores combine small effects of individual GWAS risk variants across the genome to improve personalized risk prediction. Several polygenic scores have now been developed that exhibit sufficiently large effects to be considered clinically actionable. However, their clinical use is limited by their partial transferability across ancestries and a lack of validated models that combine polygenic, monogenic, family history and clinical risk factors. Moreover, prospective studies are still needed to demonstrate the clinical utility and cost-effectiveness of polygenic scores in clinical practice. Here, we discuss evolving methods for developing polygenic scores, best practices for validating and reporting their performance, and the study designs that will empower their clinical implementation. We specifically focus on the polygenic scores relevant to nephrology and other chronic, complex diseases and review their key limitations, necessary refinements and potential clinical applications.

Genome-wide studies define new genetic mechanisms of IgA vasculitis

IgA vasculitis (IgAV) is a pediatric disease with skin and systemic manifestations. Here, we conducted genome, transcriptome, and proteome-wide association studies in 2,170 IgAV cases and 5,928 controls, generated IgAV-specific maps of gene expression and splicing from blood of 255 pediatric cases, and reconstructed myeloid-specific regulatory networks to define disease master regulators modulated by the newly identified disease driver genes. We observed significant association at the HLA-DRB1 (OR=1.55, P=1.1×10-25) and fine-mapped specific amino-acid risk substitutions in DRβ1. We discovered two novel non-HLA loci: FCAR (OR=1.51, P=1.0×10-20) encoding a myeloid IgA receptor FcαR, and INPP5D (OR=1.34, P=2.2×10-9) encoding a known inhibitor of FcαR signaling. The FCAR risk locus co-localized with a cis-eQTL increasing FCAR expression; the risk alleles disrupted a PRDM1 binding motif within a myeloid enhancer of FCAR. Another risk locus was associated with a higher genetically predicted levels of plasma IL6R. The IL6R risk haplotype carried a missense variant contributing to accelerated cleavage of IL6R into a soluble form. Using systems biology approaches, we prioritized IgAV master regulators co-modulated by FCAR, INPP5D and IL6R in myeloid cells. We additionally identified 21 shared loci in a cross-phenotype analysis of IgAV with IgA nephropathy, including novel loci PAID4, WLS, and ANKRD55.

Disease-specific prioritization of non-coding GWAS variants based on chromatin accessibility

Non-protein-coding genetic variants are a major driver of the genetic risk for human disease; however, identifying which non-coding variants contribute to diseases and their mechanisms remains challenging. In silico variant prioritization methods quantify a variant's severity, but for most methods, the specific phenotype and disease context of the prediction remain poorly defined. For example, many commonly used methods provide a single, organism-wide score for each variant, while other methods summarize a variant's impact in certain tissues and/or cell types. Here, we propose a complementary disease-specific variant prioritization scheme, which is motivated by the observation that variants contributing to disease often operate through specific biological mechanisms. We combine tissue/cell-type-specific variant scores (e.g., GenoSkyline, FitCons2, DNA accessibility) into disease-specific scores with a logistic regression approach and apply it to ∼25,000 non-coding variants spanning 111 diseases. We show that this disease-specific aggregation significantly improves the association of common non-coding genetic variants with disease (average precision: 0.151, baseline = 0.09), compared with organism-wide scores (GenoCanyon, LINSIGHT, GWAVA, Eigen, CADD; average precision: 0.129, baseline = 0.09). Further on, disease similarities based on data-driven aggregation weights highlight meaningful disease groups, and it provides information about tissues and cell types that drive these similarities. We also show that so-learned similarities are complementary to genetic similarities as quantified by genetic correlation. Overall, our approach demonstrates the strengths of disease-specific variant prioritization, leads to improvement in non-coding variant prioritization, and enables interpretable models that link variants to disease via specific tissues and/or cell types.

Modeling tissue co-regulation estimates tissue-specific contributions to disease

Integrative analyses of genome-wide association studies and gene expression data have implicated many disease-critical tissues. However, co-regulation of genetic effects on gene expression across tissues impedes distinguishing biologically causal tissues from tagging tissues. In the present study, we introduce tissue co-regulation score regression (TCSC), which disentangles causal tissues from tagging tissues by regressing gene-disease association statistics (from transcriptome-wide association studies) on tissue co-regulation scores, reflecting correlations of predicted gene expression across genes and tissues. We applied TCSC to 78 diseases/traits (average n = 302,000) and gene expression prediction models for 48 GTEx tissues. TCSC identified 21 causal tissue-trait pairs at a 5% false discovery rate (FDR), including well-established findings, biologically plausible new findings (for example, aorta artery and glaucoma) and increased specificity of known tissue-trait associations (for example, subcutaneous adipose, but not visceral adipose, and high-density lipoprotein). TCSC also identified 17 causal tissue-trait covariance pairs at 5% FDR. In conclusion, TCSC is a precise method for distinguishing causal tissues from tagging tissues.

Integrative epigenomic and functional characterization assay based annotation of regulatory activity across diverse human cell types

We introduce ChromActivity, a computational framework for predicting and annotating regulatory activity across the genome through integration of multiple epigenomic maps and various functional characterization datasets. ChromActivity generates genomewide predictions of regulatory activity associated with each functional characterization dataset across many cell types based on available epigenomic data. It then for each cell type produces (1) ChromScoreHMM genome annotations based on the combinatorial and spatial patterns within these predictions and (2) ChromScore tracks of overall predicted regulatory activity. ChromActivity provides a resource for analyzing and interpreting the human regulatory genome across diverse cell types.

Genome-wide association analyses define pathogenic signaling pathways and prioritize drug targets for IgA nephropathy

IgA nephropathy (IgAN) is a progressive form of kidney disease defined by glomerular deposition of IgA. Here we performed a genome-wide association study of 10,146 kidney-biopsy-diagnosed IgAN cases and 28,751 controls across 17 international cohorts. We defined 30 genome-wide significant risk loci explaining 11% of disease risk. A total of 16 loci were new, including TNFSF4/TNFSF18, REL, CD28, PF4V1, LY86, LYN, ANXA3, TNFSF8/TNFSF15, REEP3, ZMIZ1, OVOL1/RELA, ETS1, IGH, IRF8, TNFRSF13B and FCAR. The risk loci were enriched in gene orthologs causing abnormal IgA levels when genetically manipulated in mice. We also observed a positive genetic correlation between IgAN and serum IgA levels. High polygenic score for IgAN was associated with earlier onset of kidney failure. In a comprehensive functional annotation analysis of candidate causal genes, we observed convergence of biological candidates on a common set of inflammatory signaling pathways and cytokine ligand-receptor pairs, prioritizing potential new drug targets.

Challenges Related to the Use of Next-Generation Sequencing for the Optimization of Drug Therapy

Over the last decade, next-generation sequencing (NGS) methods have become increasingly used in various areas of human genomics. In routine clinical care, their use is already implemented in oncology to profile the mutational landscape of a tumor, as well as in rare disease diagnostics. However, its utilization in pharmacogenomics is largely lacking behind. Recent population-scale genome data has revealed that human pharmacogenes carry a plethora of rare genetic variations that are not interrogated by conventional array-based profiling methods and it is estimated that these variants could explain around 30% of the genetically encoded functional pharmacogenetic variability.To interpret the impact of such variants on drug response a multitude of computational tools have been developed, but, while there have been major advancements, it remains to be shown whether their accuracy is sufficient to improve personalized pharmacogenetic recommendations in robust trials. In addition, conventional short-read sequencing methods face difficulties in the interrogation of complex pharmacogenes and high NGS test costs require stringent evaluations of cost-effectiveness to decide about reimbursement by national healthcare programs. Here, we illustrate current challenges and discuss future directions toward the clinical implementation of NGS to inform genotype-guided decision-making.

Imputation method for single-cell RNA-seq data using neural topic model

Single-cell RNA sequencing (scRNA-seq) technology studies transcriptome and cell-to-cell differences from higher single-cell resolution and different perspectives. Despite the advantage of high capture efficiency, downstream functional analysis of scRNA-seq data is made difficult by the excess of zero values (i.e., the dropout phenomenon). To effectively address this problem, we introduced scNTImpute, an imputation framework based on a neural topic model. A neural network encoder is used to extract underlying topic features of single-cell transcriptome data to infer high-quality cell similarity. At the same time, we determine which transcriptome data are affected by the dropout phenomenon according to the learning of the mixture model by the neural network. On the basis of stable cell similarity, the same gene information in other similar cells is borrowed to impute only the missing expression values. By evaluating the performance of real data, scNTImpute can accurately and efficiently identify the dropout values and imputes them accurately. In the meantime, the clustering of cell subsets is improved and the original biological information in cell clustering is solved, which is covered by technical noise. The source code for the scNTImpute module is available as open source at https://github.com/qiyueyang-7/scNTImpute.git.

Deep learning-assisted genome-wide characterization of massively parallel reporter assays

Massively parallel reporter assay (MPRA) is a high-throughput method that enables the study of the regulatory activities of tens of thousands of DNA oligonucleotides in a single experiment. While MPRA experiments have grown in popularity, their small sample sizes compared to the scale of the human genome limits our understanding of the regulatory effects they detect. To address this, we develop a deep learning model, MpraNet, to distinguish potential MPRA targets from the background genome. This model achieves high discriminative performance (AUROC = 0.85) at differentiating MPRA positives from a set of control variants that mimic the background genome when applied to the lymphoblastoid cell line. We observe that existing functional scores represent very distinct functional effects, and most of them fail to characterize the regulatory effect that MPRA detects. Using MpraNet, we predict potential MPRA functional variants across the genome and identify the distributions of MPRA effect relative to other characteristics of genetic variation, including allele frequency, alternative functional annotations specified by FAVOR, and phenome-wide associations. We also observed that the predicted MPRA positives are not uniformly distributed across the genome; instead, they are clumped together in active regions comprising 9.95% of the genome and inactive regions comprising 89.07% of the genome. Furthermore, we propose our model as a screen to filter MPRA experiment candidates at genome-wide scale, enabling future experiments to be more cost-efficient by increasing precision relative to that observed from previous MPRAs.

Exploration of Tools for the Interpretation of Human Non-Coding Variants

The advent of Whole Genome Sequencing (WGS) broadened the genetic variation detection range, revealing the presence of variants even in non-coding regions of the genome, which would have been missed using targeted approaches. One of the most challenging issues in WGS analysis regards the interpretation of annotated variants. This review focuses on tools suitable for the functional annotation of variants falling into non-coding regions. It couples the description of non-coding genomic areas with the results and performance of existing tools for a functional interpretation of the effect of variants in these regions. Tools were tested in a controlled genomic scenario, representing the ground-truth and allowing us to determine software performance.

Genetic variants associated with psychiatric disorders are enriched at epigenetically active sites in lymphoid cells

Multiple psychiatric disorders have been associated with abnormalities in both the innate and adaptive immune systems. The role of these abnormalities in pathogenesis, and whether they are driven by psychiatric risk variants, remains unclear. We test for enrichment of GWAS variants associated with multiple psychiatric disorders (cross-disorder or trans-diagnostic risk), or 5 specific disorders (cis-diagnostic risk), in regulatory elements in immune cells. We use three independent epigenetic datasets representing multiple organ systems and immune cell subsets. Trans-diagnostic and cis-diagnostic risk variants (for schizophrenia and depression) are enriched at epigenetically active sites in brain tissues and in lymphoid cells, especially stimulated CD4⁺ T cells. There is no evidence for enrichment of either trans-risk or cis-risk variants for schizophrenia or depression in myeloid cells. This suggests a possible model where environmental stimuli activate T cells to unmask the effects of psychiatric risk variants, contributing to the pathogenesis of mental health disorders.

TVAR: assessing tissue-specific functional effects of non-coding variants with deep learning

MOTIVATION

Analysis of whole-genome sequencing (WGS) for genetics is still a challenge due to the lack of accurate functional annotation of non-coding variants, especially the rare ones. As eQTLs have been extensively implicated in the genetics of human diseases, we hypothesize that rare non-coding variants discovered in WGS play a regulatory role in predisposing disease risk.

RESULTS

With thousands of tissue- and cell-type-specific epigenomic features, we propose TVAR. This multi-label learning-based deep neural network predicts the functionality of non-coding variants in the genome based on eQTLs across 49 human tissues in the GTEx project. TVAR learns the relationships between high-dimensional epigenomics and eQTLs across tissues, taking the correlation among tissues into account to understand shared and tissue-specific eQTL effects. As a result, TVAR outputs tissue-specific annotations, with an average AUROC of 0.77 across these tissues. We evaluate TVAR's performance on four complex diseases (coronary artery disease, breast cancer, Type 2 diabetes and Schizophrenia), using TVAR's tissue-specific annotations, and observe its superior performance in predicting functional variants for both common and rare variants, compared with five existing state-of-the-art tools. We further evaluate TVAR's G-score, a scoring scheme across all tissues, on ClinVar, fine-mapped GWAS loci, Massive Parallel Reporter Assay (MPRA) validated variants and observe the consistently better performance of TVAR compared with other competing tools.

AVAILABILITY AND IMPLEMENTATION

The TVAR source code and its scores on the ClinVar catalog, fine mapped GWAS Loci, high confidence eQTLs from GTEx dataset, and MPRA validated functional variants are available at https://github.com/haiyang1986/TVAR.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

SNPs at SMG7 Associated with Time from Biochemical Recurrence to Prostate Cancer Death

BACKGROUND

A previous genome-wide association study identified several loci with genetic variants associated with prostate cancer survival time in two cohorts from Sweden. Whether these variants have an effect in other populations or if their effect is homogenous across the course of disease is unknown.

METHODS

These variants were genotyped in a cohort of 1,298 patients. Samples were linked with age, PSA level, Gleason score, cancer stage at surgery, and times from surgery to biochemical recurrence to death from prostate cancer. SNPs rs2702185 and rs73055188 were tested for association with prostate cancer-specific survival time using a multivariate Cox proportional hazard model. SNP rs2702185 was further tested for association with time to biochemical recurrence and time from biochemical recurrence to death with a multi-state model.

RESULTS

SNP rs2702185 at SMG7 was associated with prostate cancer-specific survival time, specifically the time from biochemical recurrence to prostate cancer death (HR, 2.5; 95% confidence interval, 1.4-4.5; P = 0.0014). Nine variants were in linkage disequilibrium (LD) with rs2702185; one, rs10737246, was found to be most likely to be functional based on LD patterns and overlap with open chromatin. Patterns of open chromatin and correlation with gene expression suggest that this SNP may affect expression of SMG7 in T cells.

CONCLUSIONS

The SNP rs2702185 at the SMG7 locus is associated with time from biochemical recurrence to prostate cancer death, and its LD partner rs10737246 is predicted to be functional.

IMPACT

These results suggest that future association studies of prostate cancer survival should consider various intervals over the course of disease.

A multi-dimensional integrative scoring framework for predicting functional variants in the human genome

Attempts to identify and prioritize functional DNA elements in coding and non-coding regions, particularly through use of in silico functional annotation data, continue to increase in popularity. However, specific functional roles can vary widely from one variant to another, making it challenging to summarize different aspects of variant function with a one-dimensional rating. Here we propose multi-dimensional annotation-class integrative estimation (MACIE), an unsupervised multivariate mixed-model framework capable of integrating annotations of diverse origin to assess multi-dimensional functional roles for both coding and non-coding variants. Unlike existing one-dimensional scoring methods, MACIE views variant functionality as a composite attribute encompassing multiple characteristics and estimates the joint posterior functional probabilities of each genomic position. This estimate offers more comprehensive and interpretable information in the presence of multiple aspects of functionality. Applied to a variety of independent coding and non-coding datasets, MACIE demonstrates powerful and robust performance in discriminating between functional and non-functional variants. We also show an application of MACIE to fine-mapping and heritability enrichment analysis by using the lipids GWAS summary statistics data from the European Network for Genetic and Genomic Epidemiology Consortium.

Genome-wide association analyses identify new Brugada syndrome risk loci and highlight a new mechanism of sodium channel regulation in disease susceptibility

Brugada syndrome (BrS) is a cardiac arrhythmia disorder associated with sudden death in young adults. With the exception of SCN5A, encoding the cardiac sodium channel Na_V1.5, susceptibility genes remain largely unknown. Here we performed a genome-wide association meta-analysis comprising 2,820 unrelated cases with BrS and 10,001 controls, and identified 21 association signals at 12 loci (10 new). Single nucleotide polymorphism (SNP)-heritability estimates indicate a strong polygenic influence. Polygenic risk score analyses based on the 21 susceptibility variants demonstrate varying cumulative contribution of common risk alleles among different patient subgroups, as well as genetic associations with cardiac electrical traits and disorders in the general population. The predominance of cardiac transcription factor loci indicates that transcriptional regulation is a key feature of BrS pathogenesis. Furthermore, functional studies conducted on MAPRE2, encoding the microtubule plus-end binding protein EB2, point to microtubule-related trafficking effects on Na_V1.5 expression as a new underlying molecular mechanism. Taken together, these findings broaden our understanding of the genetic architecture of BrS and provide new insights into its molecular underpinnings.

Cobolt: integrative analysis of multimodal single-cell sequencing data

A growing number of single-cell sequencing platforms enable joint profiling of multiple omics from the same cells. We present Cobolt, a novel method that not only allows for analyzing the data from joint-modality platforms, but provides a coherent framework for the integration of multiple datasets measured on different modalities. We demonstrate its performance on multi-modality data of gene expression and chromatin accessibility and illustrate the integration abilities of Cobolt by jointly analyzing this multi-modality data with single-cell RNA-seq and ATAC-seq datasets.

Soluble Urokinase Plasminogen Activator Receptor: Genetic Variation and Cardiovascular Disease Risk in Black Adults

BACKGROUND

suPAR (Soluble urokinase plasminogen activator receptor) has emerged as an important biomarker of coagulation, inflammation, and cardiovascular disease (CVD) risk. The contribution of suPAR to CVD risk and its genetic influence in Black populations have not been evaluated.

METHODS

We measured suPAR in 3492 Black adults from the prospective, community-based JHS (Jackson Heart Study). Cross-sectional associations of suPAR with lifestyle and CVD risk factors were assessed, whole-genome sequence data were used to evaluate genetic associations of suPAR, and relationships of suPAR with incident CVD outcomes and overall mortality were estimated over follow-up.

RESULTS

In Cox models adjusted for traditional CVD risk factors, estimated glomerular filtration rate, and CRP (C-reactive protein), each 1-SD higher suPAR was associated with a 21% to 31% increased risk of incident coronary heart disease, heart failure, stroke, and mortality. In the genome-wide association study, 2 missense (rs399145 encoding p.Thr86Ala, rs4760 encoding p.Phe272Leu) and 2 noncoding regulatory variants (rs73935023 within an enhancer element and rs4251805 within the promoter) of PLAUR on chromosome 19 were each independently associated with suPAR and together explained 14% of suPAR phenotypic variation. The allele frequencies of each of the four suPAR-associated genetic variants differ considerably across African and European populations. We further show that PLAUR rs73935023 can alter transcriptional activity in vitro. We did not find any association between genetically determined suPAR and CVD in JHS or a larger electronic medical record-based analyses of Blacks or Whites.

CONCLUSIONS

Our results demonstrate the importance of ancestry-differentiated genetic variation on suPAR levels and indicate suPAR is a CVD biomarker in Black adults.

Incorporating functional priors improves polygenic prediction accuracy in UK Biobank and 23andMe data sets

Polygenic risk prediction is a widely investigated topic because of its promising clinical applications. Genetic variants in functional regions of the genome are enriched for complex trait heritability. Here, we introduce a method for polygenic prediction, LDpred-funct, that leverages trait-specific functional priors to increase prediction accuracy. We fit priors using the recently developed baseline-LD model, including coding, conserved, regulatory, and LD-related annotations. We analytically estimate posterior mean causal effect sizes and then use cross-validation to regularize these estimates, improving prediction accuracy for sparse architectures. We applied LDpred-funct to predict 21 highly heritable traits in the UK Biobank (avg N = 373 K as training data). LDpred-funct attained a +4.6% relative improvement in average prediction accuracy (avg prediction R2 = 0.144; highest R2 = 0.413 for height) compared to SBayesR (the best method that does not incorporate functional information). For height, meta-analyzing training data from UK Biobank and 23andMe cohorts (N = 1107 K) increased prediction R2 to 0.431. Our results show that incorporating functional priors improves polygenic prediction accuracy, consistent with the functional architecture of complex traits.

Prioritization of regulatory variants with tissue-specific function in the non-coding regions of human genome

Understanding the functional consequences of genetic variation in the non-coding regions of the human genome remains a challenge. We introduce h ere a computational tool, TURF, to prioritize regulatory variants with tissue-specific function by leveraging evidence from functional genomics experiments, including over 3000 functional genomics datasets from the ENCODE project provided in the RegulomeDB database. TURF is able to generate prediction scores at both organism and tissue/organ-specific levels for any non-coding variant on the genome. We present that TURF has an overall top performance in prediction by using validated variants from MPRA experiments. We also demonstrate how TURF can pick out the regulatory variants with tissue-specific function over a candidate list from associate studies. Furthermore, we found that various GWAS traits showed the enrichment of regulatory variants predicted by TURF scores in the trait-relevant organs, which indicates that these variants can be a valuable source for future studies.

VannoPortal: multiscale functional annotation of human genetic variants for interrogating molecular mechanism of traits and diseases

Interpreting the molecular mechanism of genomic variations and their causal relationship with diseases/traits are important and challenging problems in the human genetic study. To provide comprehensive and context-specific variant annotations for biologists and clinicians, here, by systematically integrating over 4TB genomic/epigenomic profiles and frequently-used annotation databases from various biological domains, we develop a variant annotation database, called VannoPortal. In general, the database has following major features: (i) systematically integrates 40 genome-wide variant annotations and prediction scores regarding allele frequency, linkage disequilibrium, evolutionary signature, disease/trait association, tissue/cell type-specific epigenome, base-wise functional prediction, allelic imbalance and pathogenicity; (ii) equips with our recent novel index system and parallel random-sweep searching algorithms for efficient management of backend databases and information extraction; (iii) greatly expands context-dependent variant annotation to incorporate large-scale epigenomic maps and regulatory profiles (such as EpiMap) across over 33 tissue/cell types; (iv) compiles many genome-scale base-wise prediction scores for regulatory/pathogenic variant classification beyond protein-coding region; (v) enables fast retrieval and direct comparison of functional evidence among linked variants using highly interactive web panel in addition to plain table; (vi) introduces many visualization functions for more efficient identification and interpretation of functional variants in single web page. VannoPortal is freely available at http://mulinlab.org/vportal.

Learning interpretable cellular and gene signature embeddings from single-cell transcriptomic data

The advent of single-cell RNA sequencing (scRNA-seq) technologies has revolutionized transcriptomic studies. However, large-scale integrative analysis of scRNA-seq data remains a challenge largely due to unwanted batch effects and the limited transferabilty, interpretability, and scalability of the existing computational methods. We present single-cell Embedded Topic Model (scETM). Our key contribution is the utilization of a transferable neural-network-based encoder while having an interpretable linear decoder via a matrix tri-factorization. In particular, scETM simultaneously learns an encoder network to infer cell type mixture and a set of highly interpretable gene embeddings, topic embeddings, and batch-effect linear intercepts from multiple scRNA-seq datasets. scETM is scalable to over 106 cells and confers remarkable cross-tissue and cross-species zero-shot transfer-learning performance. Using gene set enrichment analysis, we find that scETM-learned topics are enriched in biologically meaningful and disease-related pathways. Lastly, scETM enables the incorporation of known gene sets into the gene embeddings, thereby directly learning the associations between pathways and topics via the topic embeddings.

Integrative analysis of scRNA-seq and GWAS data pinpoints periportal hepatocytes as the relevant liver cell types for blood lipids

Liver, a heterogeneous tissue consisting of various cell types, is known to be relevant for blood lipid traits. By integrating summary statistics from genome-wide association studies (GWAS) of lipid traits and single-cell transcriptome data of the liver, we sought to identify specific cell types in the liver that were most relevant for blood lipid levels. We conducted differential expression analyses for 40 cell types from human and mouse livers in order to construct the cell-type specifically expressed gene sets, which we refer to as construction of the liver cell-type specifically expressed gene sets (CT-SEGS). Under the assumption that CT-SEGS represented specific functions of each cell type, we applied stratified linkage disequilibrium score regression to determine cell types that were most relevant for complex traits and diseases. We first confirmed the validity of this method (of delineating functionally relevant cell types) by identifying the immune cell types as relevant for autoimmune diseases. We further showed that lipid GWAS signals were enriched in the human and mouse periportal hepatocytes. Our results provide important information to facilitate future cellular studies of the metabolic mechanism affecting blood lipid levels.

A semi-supervised model to predict regulatory effects of genetic variants at single nucleotide resolution using massively parallel reporter assays

MOTIVATION

Predicting regulatory effects of genetic variants is a challenging but important problem in functional genomics. Given the relatively low sensitivity of functional assays, and the pervasiveness of class imbalance in functional genomic data, popular statistical prediction models can sharply underestimate the probability of a regulatory effect. We describe here the presence-only model (PO-EN), a type of semi-supervised model, to predict regulatory effects of genetic variants at sequence-level resolution in a context of interest by integrating a large number of epigenetic features and massively parallel reporter assays (MPRAs).

RESULTS

Using experimental data from a variety of MPRAs we show that the presence-only model produces better calibrated predicted probabilities and has increased accuracy relative to state-of-the-art prediction models. Furthermore, we show that the predictions based on pre-trained PO-EN models are useful for prioritizing functional variants among candidate eQTLs and significant SNPs at GWAS loci. In particular, for the costimulatory locus, associated with multiple autoimmune diseases, we show evidence of a regulatory variant residing in an enhancer 24.4 kb downstream of CTLA4, with evidence from capture Hi-C of interaction with CTLA4. Furthermore, the risk allele of the regulatory variant is on the same risk increasing haplotype as a functional coding variant in exon 1 of CTLA4, suggesting that the regulatory variant acts jointly with the coding variant leading to increased risk to disease.

AVAILABILITY

The presence-only model is implemented in the R package 'PO.EN', freely available on CRAN. A vignette describing a detailed demonstration of using the proposed PO-EN model can be found on github at https://github.com/Iuliana-Ionita-Laza/PO.EN/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Medical Records-Based Genetic Studies of the Complement System

BACKGROUND

Genetic variants in complement genes have been associated with a wide range of human disease states, but well-powered genetic association studies of complement activation have not been performed in large multiethnic cohorts.

METHODS

We performed medical records-based genome-wide and phenome-wide association studies for plasma C3 and C4 levels among participants of the Electronic Medical Records and Genomics (eMERGE) network.

RESULTS

In a GWAS for C3 levels in 3949 individuals, we detected two genome-wide significant loci: chr.1q31.3 (CFH locus; rs3753396-A; β=0.20; 95% CI, 0.14 to 0.25; P=1.52x10-11) and chr.19p13.3 (C3 locus; rs11569470-G; β=0.19; 95% CI, 0.13 to 0.24; P=1.29x10-8). These two loci explained approximately 2% of variance in C3 levels. GWAS for C4 levels involved 3998 individuals and revealed a genome-wide significant locus at chr.6p21.32 (C4 locus; rs3135353-C; β=0.40; 95% CI, 0.34 to 0.45; P=4.58x10-35). This locus explained approximately 13% of variance in C4 levels. The multiallelic copy number variant analysis defined two structural genomic C4 variants with large effect on blood C4 levels: C4-BS (β=-0.36; 95% CI, -0.42 to -0.30; P=2.98x10-22) and C4-AL-BS (β=0.25; 95% CI, 0.21 to 0.29; P=8.11x10-23). Overall, C4 levels were strongly correlated with copy numbers of C4A and C4B genes. In comprehensive phenome-wide association studies involving 102,138 eMERGE participants, we cataloged a full spectrum of autoimmune, cardiometabolic, and kidney diseases genetically related to systemic complement activation.

CONCLUSIONS

We discovered genetic determinants of plasma C3 and C4 levels using eMERGE genomic data linked to electronic medical records. Genetic variants regulating C3 and C4 levels have large effects and multiple clinical correlations across the spectrum of complement-related diseases in humans.

Multitrait GWAS to connect disease variants and biological mechanisms

Genome-wide association studies (GWASs) have uncovered a wealth of associations between common variants and human phenotypes. Here, we present an integrative analysis of GWAS summary statistics from 36 phenotypes to decipher multitrait genetic architecture and its link with biological mechanisms. Our framework incorporates multitrait association mapping along with an investigation of the breakdown of genetic associations into clusters of variants harboring similar multitrait association profiles. Focusing on two subsets of immunity and metabolism phenotypes, we then demonstrate how genetic variants within clusters can be mapped to biological pathways and disease mechanisms. Finally, for the metabolism set, we investigate the link between gene cluster assignment and the success of drug targets in randomized controlled trials.

A semi-supervised deep learning approach for predicting the functional effects of genomic non-coding variations

Background

Understanding the functional effects of non-coding variants is important as they are often associated with gene-expression alteration and disease development. Over the past few years, many computational tools have been developed to predict their functional impact. However, the intrinsic difficulty in dealing with the scarcity of data leads to the necessity to further improve the algorithms. In this work, we propose a novel method, employing a semi-supervised deep-learning model with pseudo labels, which takes advantage of learning from both experimentally annotated and unannotated data.

Results

We prepared known functional non-coding variants with histone marks, DNA accessibility, and sequence context in GM12878, HepG2, and K562 cell lines. Applying our method to the dataset demonstrated its outstanding performance, compared with that of existing tools. Our results also indicated that the semi-supervised model with pseudo labels achieves higher predictive performance than the supervised model without pseudo labels. Interestingly, a model trained with the data in a certain cell line is unlikely to succeed in other cell lines, which implies the cell-type-specific nature of the non-coding variants. Remarkably, we found that DNA accessibility significantly contributes to the functional consequence of variants, which suggests the importance of open chromatin conformation prior to establishing the interaction of non-coding variants with gene regulation.

Conclusions

The semi-supervised deep learning model coupled with pseudo labeling has advantages in studying with limited datasets, which is not unusual in biology. Our study provides an effective approach in finding non-coding mutations potentially associated with various biological phenomena, including human diseases.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-021-03999-8.

Cell Type-Specific Annotation and Fine Mapping of Variants Associated With Brain Disorders

Common genetic variants confer susceptibility to a large number of complex brain disorders. Given that such variants predominantly localize in non-coding regions of the human genome, there is a significant challenge to predict and characterize their functional consequences. More importantly, most available computational methods, generally defined as context-free methods, output prediction scores regarding the functionality of genetic variants irrespective of the context, i.e., the tissue or cell-type affected by a disease, limiting the ability to predict the functional consequences of common variants on brain disorders. In this study, we introduce a comparative multi-step pipeline to investigate the relative effectiveness of context-specific and context-free approaches to prioritize disease causal variants. As an experimental case, we focused on schizophrenia (SCZ), a debilitating neuropsychiatric disease for which a large number of susceptibility variants is identified from genome-wide association studies. We tested over two dozen available methods and examined potential associations between the cell/tissue-specific mapping scores and open chromatin accessibility, and provided a prioritized map of SCZ risk loci for in vitro or in-vivo functional analysis. We found extensive differences between context-free and tissue-specific approaches and showed how they may play complementary roles. As a proof of concept, we found a few sets of genes, through a consensus mapping of both categories, including FURIN to be among the top hits. We showed that the genetic variants in this gene and related genes collectively dysregulate gene expression patterns in stem cell-derived neurons and characterize SCZ phenotypic manifestations, while genes which were not shared among highly prioritized candidates in both approaches did not demonstrate such characteristics. In conclusion, by combining context-free and tissue-specific predictions, our pipeline enables prioritization of the most likely disease-causal common variants in complex brain disorders.

Probabilistic topic modelling in food spoilage analysis: A case study with Atlantic salmon (Salmo salar)

Probabilistic topic modelling is frequently used in machine learning and statistical analysis for extracting latent information from complex datasets. Despite being closely associated with natural language processing and text mining, these methods possess several properties that make them particularly attractive in metabolomics applications where the applicability of traditional multivariate statistics tends to be limited. The aim of the study was thus to introduce probabilistic topic modelling - more specifically, Latent Dirichlet Allocation (LDA) - in a novel experimental context: volatilome-based (sea) food spoilage characterization. This was realized as a case study, focusing on modelling the spoilage of Atlantic salmon (Salmo salar) at 4 °C under different gaseous atmospheres (% CO₂/O₂/N₂): 0/0/100 (A), air (B), 60/0/40 (C) or 60/40/0 (D). First, an exploratory analysis was performed to optimize the model tunings and to consequently model salmon spoilage under 100% N₂ (A). Based on the obtained results, a systematic spoilage characterization protocol was established and used for identifying potential volatile spoilage indicators under all tested storage conditions. In conclusion, LDA could be used for extracting sets of underlying VOC profiles and identifying those signifying salmon spoilage, giving rise to an extensive discussion regarding the key points associated with model tuning and/or spoilage analysis. The identified compounds were well in accordance with a previously established approach based on partial least squares regression analysis (PLS). Overall, the outcomes of the study not only reflect the promising potential of LDA in spoilage characterization, but also provide several new insights into the development of data-driven methods for food quality analysis.

An Imperative Need for Further Genetic Studies of Alopecia Areata

Human genetic studies of diseases that are multifactorial and prevalent have generated a wealth of knowledge about the genetic architecture of chronic diseases. Generalizable attributes are shaping the development of models to explain how the human genome influences our health and can be leveraged to improve it. Importantly, both rare and common genetic variants contribute to disease risk and provide complementary information. Although initial genetic studies of alopecia areata have yielded insight with high clinical impact, there remains a number of important unanswered questions pertaining to disease biology and patient care that could be addressed by further genetic investigations.

Ultrafast and scalable variant annotation and prioritization with big functional genomics data

The advances of large-scale genomics studies have enabled compilation of cell type–specific, genome-wide DNA functional elements at high resolution. With the growing volume of functional annotation data and sequencing variants, existing variant annotation algorithms lack the efficiency and scalability to process big genomic data, particularly when annotating whole-genome sequencing variants against a huge database with billions of genomic features. Here, we develop VarNote to rapidly annotate genome-scale variants in large and complex functional annotation resources. Equipped with a novel index system and a parallel random-sweep searching algorithm, VarNote shows substantial performance improvements (two to three orders of magnitude) over existing algorithms at different scales. It supports both region-based and allele-specific annotations and introduces advanced functions for the flexible extraction of annotations. By integrating massive base-wise and context-dependent annotations in the VarNote framework, we introduce three efficient and accurate pipelines to prioritize the causal regulatory variants for common diseases, Mendelian disorders, and cancers.

Dynamic incorporation of multiple in silico functional annotations empowers rare variant association analysis of large whole-genome sequencing studies at scale

Large-scale whole-genome sequencing studies have enabled the analysis of rare variants (RVs) associated with complex phenotypes. Commonly used RV association tests have limited scope to leverage variant functions. We propose STAAR (variant-set test for association using annotation information), a scalable and powerful RV association test method that effectively incorporates both variant categories and multiple complementary annotations using a dynamic weighting scheme. For the latter, we introduce 'annotation principal components', multidimensional summaries of in silico variant annotations. STAAR accounts for population structure and relatedness and is scalable for analyzing very large cohort and biobank whole-genome sequencing studies of continuous and dichotomous traits. We applied STAAR to identify RVs associated with four lipid traits in 12,316 discovery and 17,822 replication samples from the Trans-Omics for Precision Medicine Program. We discovered and replicated new RV associations, including disruptive missense RVs of NPC1L1 and an intergenic region near APOC1P1 associated with low-density lipoprotein cholesterol.

Trans-ancestral dissection of urate- and gout-associated major loci SLC2A9 and ABCG2 reveals primate-specific regulatory effects

Gout is a complex inflammatory arthritis affecting ~20% of people with an elevated serum urate level (hyperuricemia). Gout and hyperuricemia are essentially specific to humans and other higher primates, with varied prevalence across ancestral groups. SLC2A9 and ABCG2 are major loci associated with both urate and gout in multiple ancestral groups. However, fine mapping has been challenging due to extensive linkage disequilibrium underlying the associated regions. We used trans-ancestral fine mapping integrated with primate-specific genomic information to address this challenge. Trans-ancestral meta-analyses of GWAS cohorts of either European (EUR) or East Asian (EAS) ancestry resulted in single-variant resolution mappings for SLC2A9 (rs3775948 for urate and rs4697701 for gout) and ABCG2 (rs2622621 for gout). Tests of colocalization of variants in both urate and gout suggested existence of a shared candidate causal variant for SLC2A9 only in EUR and for ABCG2 only in EAS. The fine-mapped gout variant rs4697701 was within an ancient enhancer, whereas rs2622621 was within a primate-specific transposable element, both supported by functional evidence from the Roadmap Epigenomics project in human primary tissues relevant to urate and gout. Additional primate-specific elements were found near both loci and those adjacent to SLC2A9 overlapped with known statistical epistatic interactions associated with urate as well as multiple super-enhancers identified in urate-relevant tissues. We conclude that by leveraging ancestral differences trans-ancestral fine mapping has identified ancestral and functional variants for SLC2A9 or ABCG2 with primate-specific regulatory effects on urate and gout.

Statistical Methods in Genome-Wide Association Studies

Since the initial success of genome-wide association studies (GWAS) in 2005, tens of thousands of genetic variants have been identified for hundreds of human diseases and traits. In a GWAS, genotype information at up to millions of genetic markers is collected from up to hundreds of thousands of individuals, together with their phenotype information. Several scientific goals can be accomplished through the analysis of GWAS data, including the identification of variants, genes, and pathways associated with diseases and traits of interest; the inference of the genetic architecture of these traits; and the development of genetic risk prediction models. In this review, we provide an overview of the statistical challenges in achieving these goals and recent progress in statistical methodology to address these challenges.

Modeling neuronal consequences of autism-associated gene regulatory variants with human induced pluripotent stem cells

Genetic factors contribute to the development of autism spectrum disorder (ASD), and although non-protein-coding regions of the genome are being increasingly implicated in ASD, the functional consequences of these variants remain largely uncharacterized. Induced pluripotent stem cells (iPSCs) enable the production of personalized neurons that are genetically matched to people with ASD and can therefore be used to directly test the effects of genomic variation on neuronal gene expression, synapse function, and connectivity. The combined use of human pluripotent stem cells with genome editing to introduce or correct specific variants has proved to be a powerful approach for exploring the functional consequences of ASD-associated variants in protein-coding genes and, more recently, long non-coding RNAs (lncRNAs). Here, we review recent studies that implicate lncRNAs, other non-coding mutations, and regulatory variants in ASD susceptibility. We also discuss experimental design considerations for using iPSCs and genome editing to study the role of the non-protein-coding genome in ASD.

Annotations capturing cell type-specific TF binding explain a large fraction of disease heritability

Regulatory variation plays a major role in complex disease and that cell type-specific binding of transcription factors (TF) is critical to gene regulation. However, assessing the contribution of genetic variation in TF-binding sites to disease heritability is challenging, as binding is often cell type-specific and annotations from directly measured TF binding are not currently available for most cell type-TF pairs. We investigate approaches to annotate TF binding, including directly measured chromatin data and sequence-based predictions. We find that TF-binding annotations constructed by intersecting sequence-based TF-binding predictions with cell type-specific chromatin data explain a large fraction of heritability across a broad set of diseases and corresponding cell types; this strategy of constructing annotations addresses both the limitation that identical sequences may be bound or unbound depending on surrounding chromatin context and the limitation that sequence-based predictions are generally not cell type-specific. We partitioned the heritability of 49 diseases and complex traits using stratified linkage disequilibrium (LD) score regression with the baseline-LD model (which is not cell type-specific) plus the new annotations. We determined that 100 bp windows around MotifMap sequenced-based TF-binding predictions intersected with a union of six cell type-specific chromatin marks (imputed using ChromImpute) performed best, with an 58% increase in heritability enrichment compared to the chromatin marks alone (11.6× vs. 7.3×, P = 9 × 10-14 for difference) and a 20% increase in cell type-specific signal conditional on annotations from the baseline-LD model (P = 8 × 10-11 for difference). Our results show that TF-binding annotations explain substantial disease heritability and can help refine genome-wide association signals.

Loss-of-function tolerance of enhancers in the human genome

Previous studies have surveyed the potential impact of loss-of-function (LoF) variants and identified LoF-tolerant protein-coding genes. However, the tolerance of human genomes to losing enhancers has not yet been evaluated. Here we present the catalog of LoF-tolerant enhancers using structural variants from whole-genome sequences. Using a conservative approach, we estimate that individual human genomes possess at least 28 LoF-tolerant enhancers on average. We assessed the properties of LoF-tolerant enhancers in a unified regulatory network constructed by integrating tissue-specific enhancers and gene-gene interactions. We find that LoF-tolerant enhancers tend to be more tissue-specific and regulate fewer and more dispensable genes relative to other enhancers. They are enriched in immune-related cells while enhancers with low LoF-tolerance are enriched in kidney and brain/neuronal stem cells. We developed a supervised learning approach to predict the LoF-tolerance of all enhancers, which achieved an area under the receiver operating characteristics curve (AUROC) of 98%. We predict 3,519 more enhancers would be likely tolerant to LoF and 129 enhancers that would have low LoF-tolerance. Our predictions are supported by a known set of disease enhancers and novel deletions from PacBio sequencing. The LoF-tolerance scores provided here will serve as an important reference for disease studies.

Enhancers are elements where transcription factors bind and regulate the expression of protein-coding genes. Although multiple previous studies have focused on which genes can tolerate loss-of-function (LoF), none has systematically evaluated the tolerance of all enhancers in the human genome to LoF. Individual studies have shown a broad range of phenotypic effects of enhancer LoF. The phenotypic effects of enhancer LoF likely fall into a spectrum where deletion of LoF-tolerant enhancers would not elicit substantial phenotypic impact, while some enhancers are likely to cause fitness defects when deleted. Here we report a systematic computational approach that uses machine learning and properties of enhancers in a unified human regulatory network with tissue-specific annotations to predict the LoF-tolerance of all enhancers identified in the human genome. The LoF-tolerance scores of enhancers provided in this study can significantly facilitate the interpretation and prioritization of non-coding sequence variants for disease and functional studies.

Co-localization between Sequence Constraint and Epigenomic Information Improves Interpretation of Whole-Genome Sequencing Data

The identification of functional regions in the noncoding human genome is difficult but critical in order to gain understanding of the role noncoding variation plays in gene regulation in human health and disease. We describe here a co-localization approach that aims to identify constrained sequences that co-localize with tissue- or cell-type-specific regulatory regions, and we show that the resulting score is particularly well suited for the identification of rare regulatory variants. For 127 tissues and cell types in the ENCODE/Roadmap Epigenomics Project, we provide catalogs of putative tissue- or cell-type-specific regulatory regions under sequence constraint. We use the newly developed co-localization score for brain tissues to score de novo mutations in whole genomes from 1,902 individuals affected with autism spectrum disorder (ASD) and their unaffected siblings in the Simons Simplex Collection. We show that noncoding de novo mutations near genes co-expressed in midfetal brain with high confidence ASD risk genes, and near FMRP gene targets are more likely to be in co-localized regions if they occur in ASD probands versus in their unaffected siblings. We also observed a similar enrichment for mutations near lincRNAs, previously shown to co-express with ASD risk genes. Additionally, we provide strong evidence that prioritized de novo mutations in autism probands point to a small set of well-known ASD genes, the disruption of which produces relevant mouse phenotypes such as abnormal social investigation and abnormal discrimination/associative learning, unlike the de novo mutations in unaffected siblings. The genome-wide co-localization results are available online.

The genetic architecture of membranous nephropathy and its potential to improve non-invasive diagnosis

Membranous Nephropathy (MN) is a rare autoimmune cause of kidney failure. Here we report a genome-wide association study (GWAS) for primary MN in 3,782 cases and 9,038 controls of East Asian and European ancestries. We discover two previously unreported loci, NFKB1 (rs230540, OR = 1.25, P = 3.4 × 10-12) and IRF4 (rs9405192, OR = 1.29, P = 1.4 × 10-14), fine-map the PLA2R1 locus (rs17831251, OR = 2.25, P = 4.7 × 10-103) and report ancestry-specific effects of three classical HLA alleles: DRB1*1501 in East Asians (OR = 3.81, P = 2.0 × 10-49), DQA1*0501 in Europeans (OR = 2.88, P = 5.7 × 10-93), and DRB1*0301 in both ethnicities (OR = 3.50, P = 9.2 × 10-23 and OR = 3.39, P = 5.2 × 10-82, respectively). GWAS loci explain 32% of disease risk in East Asians and 25% in Europeans, and correctly re-classify 20-37% of the cases in validation cohorts that are antibody-negative by the serum anti-PLA2R ELISA diagnostic test. Our findings highlight an unusual genetic architecture of MN, with four loci and their interactions accounting for nearly one-third of the disease risk.

Whole-Genome k-mer Topic Modeling AssociatesBacterial Families

Alignment-free k-mer-based algorithms in whole genome sequence comparisons remainan ongoing challenge. Here, we explore the possibility to use Topic Modeling for organismwhole-genome comparisons. We analyzed 30 complete genomes from three bacterial families bytopic modeling. For this, each genome was considered as a document and 13-mer nucleotiderepresentations as words. Latent Dirichlet allocation was used as the probabilistic modeling of thecorpus. We where able to identify the topic distribution among analyzed genomes, which is highlyconsistent with traditional hierarchical classification. It is possible that topic modeling may be appliedto establish relationships between genome's composition and biological phenomena.

Novel and Known Gene-Smoking Interactions With cIMT Identified as Potential Drivers for Atherosclerosis Risk in West-African Populations of the AWI-Gen Study

Introduction

Atherosclerosis is a key contributor to the burden of cardiovascular diseases (CVDs) and many epidemiological studies have reported on the effect of smoking on carotid intima-media thickness (cIMT) and its subsequent effect on CVD risk. Gene-environment interaction studies have contributed towards understanding some of the missing heritability of genome-wide association studies. Gene-smoking interactions on cIMT have been studied in non-African populations (European, Latino-American, and African American) but no comparable African research has been reported. Our aim was to investigate smoking-SNP interactions on cIMT in two West African populations by genome-wide analysis.

Materials and methods

Only male participants from Burkina Faso (Nanoro = 993) and Ghana (Navrongo = 783) were included, as smoking was extremely rare among women. Phenotype and genotype data underwent stringent QC and genotype imputation was performed using the Sanger African Imputation Panel. Smoking prevalence among men was 13.3% in Nanoro and 42.5% in Navrongo. We analyzed gene-smoking interactions with PLINK after adjusting for covariates: age and 6 PCs (Model 1); age, BMI, blood pressure, fasting glucose, cholesterol levels, MVPA, and 6 PCs (Model 2). All analyses were performed at site level and for the combined data set.

Results

In Nanoro, we identified new gene-smoking interaction variants for cIMT within the previously described RCBTB1 region (rs112017404, rs144170770, and rs4941649) (Model 1: p = 1.35E-07; Model 2: p = 3.08E-08). In the combined sample, two novel intergenic interacting variants were identified, rs1192824 in the regulatory region of TBC1D8 (p = 5.90E-09) and rs77461169 (p = 4.48E-06) located in an upstream region of open chromatin. In silico functional analysis suggests the involvement of genes implicated in biological processes related to cell or biological adhesion and regulatory processes in gene-smoking interactions with cIMT (as evidenced by chromatin interactions and eQTLs).

Discussion

This is the first gene-smoking interaction study for cIMT, as a risk factor for atherosclerosis, in sub-Saharan African populations. In addition to replicating previously known signals for RCBTB1, we identified two novel genomic regions (TBC1D8, near BCHE) involved in this gene-environment interaction.

Methodological challenges in constructing DNA methylation risk scores

Polygenic approaches often access more variance of complex traits than is possible by single variant approaches. For genotype data, genetic risk scores (GRS) are widely used for risk prediction as well as in association and interaction studies. Recently, interest has been growing in transferring GRS approaches to DNA methylation data (methylation risk scores, MRS), which can be used 1) as biomarkers for environmental exposures, 2) in association analyses in which single CpG sites do not achieve significance, 3) as dimension reduction approach in interaction and mediation analyses, and 4) to predict individual risks of disease or treatment success. Most GRS approaches can directly be transferred to methylation data. However, since methylation data is more sensitive to confounding, e.g. by age and tissue, it is more complex to find appropriate external weights. In this review, we will outline the adaption of current GRS approaches to methylation data and highlight occurring challenges.

Prediction of apoptosis protein subcellular location based on position-specific scoring matrix and isometric mapping algorithm

Apoptosis proteins are related to many diseases. Obtaining the subcellular localization information of apoptosis proteins is helpful to understand the mechanism of diseases and to develop new drugs. At present, the researchers mainly focus on the primary protein sequences, so there is still room for improvement in the prediction accuracy of the subcellular localization of apoptosis proteins. In this paper, a new method named ERT-ECT-PSSM-IS is proposed to predict apoptosis proteins based on the position-specific scoring matrix (PSSM). First, the local and global features of different directions are extracted by evolutionary row transformation (ERT) and cross-covariance of evolutionary column transformation (ECT) based on PSSM (ERT-ECT-PSSM). Second, an improved isometric mapping algorithm (I-SMA) is used to eliminate redundant features. Finally, we adopt a support vector machine (SVM) to classify our results, and the prediction accuracy is evaluated by jackknife cross-validation tests. The experimental results show that the proposed method not only extracts more abundant feature expression but also has better predictive performance and robustness for the subcellular localization of apoptosis proteins in ZD98, ZW225, and CL317 databases. Graphical abstract Framework of the proposed prediction model.

GRAM: A GeneRAlized Model to predict the molecular effect of a non-coding variant in a cell-type specific manner

There has been much effort to prioritize genomic variants with respect to their impact on "function". However, function is often not precisely defined: sometimes it is the disease association of a variant; on other occasions, it reflects a molecular effect on transcription or epigenetics. Here, we coupled multiple genomic predictors to build GRAM, a GeneRAlized Model, to predict a well-defined experimental target: the expression-modulating effect of a non-coding variant on its associated gene, in a transferable, cell-specific manner. Firstly, we performed feature engineering: using LASSO, a regularized linear model, we found transcription factor (TF) binding most predictive, especially for TFs that are hubs in the regulatory network; in contrast, evolutionary conservation, a popular feature in many other variant-impact predictors, has almost no contribution. Moreover, TF binding inferred from in vitro SELEX is as effective as that from in vivo ChIP-Seq. Second, we implemented GRAM integrating only SELEX features and expression profiles; thus, the program combines a universal regulatory score with an easily obtainable modifier reflecting the particular cell type. We benchmarked GRAM on large-scale MPRA datasets, achieving AUROC scores of 0.72 in GM12878 and 0.66 in a multi-cell line dataset. We then evaluated the performance of GRAM on targeted regions using luciferase assays in the MCF7 and K562 cell lines. We noted that changing the insertion position of the construct relative to the reporter gene gave very different results, highlighting the importance of carefully defining the exact prediction target of the model. Finally, we illustrated the utility of GRAM in fine-mapping causal variants and developed a practical software pipeline to carry this out. In particular, we demonstrated in specific examples how the pipeline could pinpoint variants that directly modulate gene expression within a larger linkage-disequilibrium block associated with a phenotype of interest (e.g., for an eQTL).

Innovation hotspots in food waste treatment, biogas, and anaerobic digestion technology: A natural language processing approach

The objective of this study is to apply natural language processing to identifying innovative technology trends related to food waste treatment, biogas, and anaerobic digestion. The methodology used involved analyzing large volumes of text data mined from 3186 patents related to these three fields. Latent Dirichlet Allocation and the perplexity method were used to identify the main topics which the patent corpora were comprised of and which technological concepts were most associated with each topic. In addition, term frequency-inverse document frequency (TF-IDF) was used to gauge the "emergingness" of certain technical concepts across the patent corpora in various years. The key results were as follows: (1) perplexity computations showed that a 20 topic models were feasible for these patent corpora; (2) topics were identified, providing an accurate picture of the patenting landscape in the analyzed fields; (3) TF-IDF analysis on unigrams, bigrams, and trigrams, supplemented with network graph analysis, revealed emerging technology trends in each year. This study has important implications for governments who need to decide where to invest resources in anaerobic food waste treatment.

Sufficient direction factor model and its application to gene expression quantitative trait loci discovery

Rapid improvement in technology has made it relatively cheap to collect genetic data, however statistical analysis of existing data is still much cheaper. Thus, secondary analysis of single-nucleotide polymorphism, SNP, data, i.e., reanalysing existing data in an effort to extract more information, is an attractive and cost-effective alternative to collecting new data. We study the relationship between gene expression and SNPs through a combination of factor analysis and dimension reduction estimation. To take advantage of the flexibility in traditional factor models where the latent factors are not required to be normal, we recommend using semiparametric sufficient dimension reduction methods in the joint estimation of the combined model. The resulting estimator is flexible and has superior performance relative to the existing estimator, which relies on additional assumptions on the latent factors. We quantify the asymptotic performance of the proposed parameter estimator and perform inference by assessing the estimation variability and by constructing confidence intervals. The new results enable us to identify, for the first time, statistically significant SNPs concerning gene-SNP relations in lung tissue from genotype-tissue expression data.

De novo pattern discovery enables robust assessment of functional consequences of non-coding variants

MOTIVATION

Given the complexity of genome regions, prioritize the functional effects of non-coding variants remains a challenge. Although several frameworks have been proposed for the evaluation of the functionality of non-coding variants, most of them used 'black boxes' methods that simplify the task as the pathogenicity/benign classification problem, which ignores the distinct regulatory mechanisms of variants and leads to less desirable performance. In this study, we developed DVAR, an unsupervised framework that leverage various biochemical and evolutionary evidence to distinguish the gene regulatory categories of variants and assess their comprehensive functional impact simultaneously.

RESULTS

DVAR performed de novo pattern discovery in high-dimensional data and identified five regulatory clusters of non-coding variants. Leveraging the new insights into the multiple functional patterns, it measures both the between-class and the within-class functional implication of the variants to achieve accurate prioritization. Compared to other two-class learning methods, it showed improved performance in identification of clinically significant variants, fine-mapped GWAS variants, eQTLs and expression-modulating variants. Moreover, it has superior performance on disease causal variants verified by genome-editing (like CRISPR-Cas9), which could provide a pre-selection strategy for genome-editing technologies across the whole genome. Finally, evaluated in BioVU and UK Biobank, two large-scale DNA biobanks linked to complete electronic health records, DVAR demonstrated its effectiveness in prioritizing non-coding variants associated with medical phenotypes.

AVAILABILITY AND IMPLEMENTATION

The C++ and Python source codes, the pre-computed DVAR-cluster labels and DVAR-scores across the whole genome are available at https://www.vumc.org/cgg/dvar.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Using topic modeling via non-negative matrix factorization to identify relationships between genetic variants and disease phenotypes: A case study of Lipoprotein(a) (LPA)

Genome-wide and phenome-wide association studies are commonly used to identify important relationships between genetic variants and phenotypes. Most studies have treated diseases as independent variables and suffered from the burden of multiple adjustment due to the large number of genetic variants and disease phenotypes. In this study, we used topic modeling via non-negative matrix factorization (NMF) for identifying associations between disease phenotypes and genetic variants. Topic modeling is an unsupervised machine learning approach that can be used to learn patterns from electronic health record data. We chose the single nucleotide polymorphism (SNP) rs10455872 in LPA as the predictor since it has been shown to be associated with increased risk of hyperlipidemia and cardiovascular diseases (CVD). Using data of 12,759 individuals with electronic health records (EHR) and linked DNA samples at Vanderbilt University Medical Center, we trained a topic model using NMF from 1,853 distinct phenotypes and identified six topics. We tested their associations with rs10455872 in LPA. Topics enriched for CVD and hyperlipidemia had positive correlations with rs10455872 (P < 0.001), replicating a previous finding. We also identified a negative correlation between LPA and a topic enriched for lung cancer (P < 0.001) which was not previously identified via phenome-wide scanning. We were able to replicate the top finding in a separate dataset. Our results demonstrate the applicability of topic modeling in exploring the relationship between genetic variants and clinical diseases.

NCBoost classifies pathogenic non-coding variants in Mendelian diseases through supervised learning on purifying selection signals in humans

State-of-the-art methods assessing pathogenic non-coding variants have mostly been characterized on common disease-associated polymorphisms, yet with modest accuracy and strong positional biases. In this study, we curated 737 high-confidence pathogenic non-coding variants associated with monogenic Mendelian diseases. In addition to interspecies conservation, a comprehensive set of recent and ongoing purifying selection signals in humans is explored, accounting for lineage-specific regulatory elements. Supervised learning using gradient tree boosting on such features achieves a high predictive performance and overcomes positional bias. NCBoost performs consistently across diverse learning and independent testing data sets and outperforms other existing reference methods.

Estimating contribution of rare non-coding variants to neuropsychiatric disorders

Owing to recent advances in DNA sequencing technology, a number of large-scale comprehensive analyses of genetic variations in protein-coding regions (i.e., whole-exome sequencing studies), have been conducted for neuropsychiatric and neurodevelopmental disorders, such as autism spectrum disorders, intellectual disability, and schizophrenia. These studies, especially those focusing on de novo (newly arising) mutations and extremely rare variants, have successfully identified previously unrecognized disease genes/mutations with a large effect size and deepen our understanding of the biology of neuropsychiatric diseases. Along with the continuously dropping sequencing cost, now the target of sequencing studies is expanding from the exome to the whole human genome. Several pioneering works have provided important insights into the contribution of rare non-coding variants to neuropsychiatric diseases. At the same time, these studies highlight need for further larger sample sizes and improvement in annotation of non-coding regulatory variants. In this review, key findings from recent studies as well as likely future directions are overviewed.

A semi-supervised approach for predicting cell-type specific functional consequences of non-coding variation using MPRAs

Predicting the functional consequences of genetic variants in non-coding regions is a challenging problem. We propose here a semi-supervised approach, GenoNet, to jointly utilize experimentally confirmed regulatory variants (labeled variants), millions of unlabeled variants genome-wide, and more than a thousand cell/tissue type specific epigenetic annotations to predict functional consequences of non-coding variants. Through the application to several experimental datasets, we demonstrate that the proposed method significantly improves prediction accuracy compared to existing functional prediction methods at the tissue/cell type level, but especially so at the organism level. Importantly, we illustrate how the GenoNet scores can help in fine-mapping at GWAS loci, and in the discovery of disease associated genes in sequencing studies. As more comprehensive lists of experimentally validated variants become available over the next few years, semi-supervised methods like GenoNet can be used to provide increasingly accurate functional predictions for variants genome-wide and across a variety of cell/tissue types.

Predicting the functional consequences of non-coding genetic variants is a challenge. Here, He et al. present GenoNet, a semi-supervised method that combines information from experimentally confirmed regulatory variants with cell type- and tissue specific annotation for function prediction.