Biology subjects

Halperin, E.

Publications and source records attributed to Halperin, E..

At least 19 recordsLinked to original sources

bioRxiv2025

Integrated ambient modeling and genetic demultiplexing of single-cell RNA+ATAC multiome experiments with Ambimux

Single cell technologies have advanced at a rapid pace, providing assays for various molecular phenotypes. Droplet-based single cell technologies, particularly those based on nuclei isolation, such as simultaneous RNA+ATAC single-cell multiome, are susceptible to exogenous ambient molecule contamination, which can increase noise in cell type-level associations. We reasoned that genotype-based sample multiplexing can provide an opportunity to infer this ambient contamination by leveraging DNA variation in sequenced reads. Thus, we developed ambimux, a likelihood-based method to estimate ambient fractions and demultiplex single-cell multiome experiments using genotype-level data. Ambimux models the ambient or nuclear probability at the read level and thus can classify empty droplets and estimate droplet-specific ambient molecule fractions in each modality. We first evaluated our method using simulated data sets across a range of parameters. We found that ambimux closely estimated the ground truth droplet contamination fractions in the RNA (MAE=0.048) and ATAC (MAE=0.042) modalities. As a result, ambimux maintained high specificity (>95%) and was able to correctly assign singlets at considerably high ambient fractions (up to 60%) for both RNA and ATAC modalities. In comparison with models that do not consider ambient contamination, these only maintained similar sensitivity levels at considerably lower ambient fractions (up to 25%). We then generated a real data set of seven visceral adipose tissue biopsies run on a single 10x Multiome channel. We ran ambimux and detected 4,986 singlets, capturing similar numbers as other methods. Then, we sought to evaluate the fidelity of the ambient fraction estimates from ambimux. We split singlets into ambient-enriched (>5% contamination in both modalities) or nuclear-enriched (<5% in both) droplets and performed gene-peak linkage analysis. Low ambient droplets resulted in more significant hits with gene-peak links enriched at the transcription start site relative to high ambient droplets, suggesting that the ambient droplets identified by ambimux hamper the identification of biologically meaningful signals. In summary, we developed a joint single-cell multiome demultiplexing method, ambimux, that accurately models and estimates ambient molecule contamination in each modality.

bioinformatics↗

bioRxiv2024

Single-cell DNA methylome and 3D genome atlas of the human subcutaneous adipose tissue

Human subcutaneous adipose tissue (SAT) contains a diverse array of cell-types; however, the epigenomic landscape among the SAT cell-types has remained elusive. Our integrative analysis of single-cell resolution DNA methylation and chromatin conformation profiles (snm3C-seq), coupled with matching RNA expression (snRNA-seq), systematically cataloged the epigenomic, 3D topology, and transcriptomic dynamics across the SAT cell-types. We discovered that the SAT CG methylation (mCG) landscape is characterized by pronounced hyper-methylation in myeloid cells and hypo-methylation in adipocytes and adipose stem and progenitor cells (ASPCs), driving nearly half of the 705,063 detected differentially methylated regions (DMRs). In addition to the enriched cell-type-specific transcription factor binding motifs, we identified TET1 and DNMT3A as plausible candidates for regulating cell-type level mCG profiles. Furthermore, we observed that global mCG profiles closely correspond to SAT lineage, which is also reflected in cell-type-specific chromosome compartmentalization. Adipocytes, in particular, display significantly more short-range chromosomal interactions, facilitating the formation of complex local 3D genomic structures that regulate downstream transcriptomic activity, including those associated with adipogenesis. Finally, we discovered that variants in cell-type level DMRs and A compartments significantly predict and are enriched for variance explained in abdominal obesity. Together, our multimodal study characterizes human SAT epigenomic landscape at the cell-type resolution and links partitioned polygenic risk of abdominal obesity to SAT epigenome.

genomics↗

bioRxiv2024

A unified model for cell-type resolution genomics from heterogeneous omics data

The vast majority of population-scale genomic datasets collected to date consist of "bulk" samples obtained from heterogeneous tissues, reflecting mixtures of different cell types. In order to facilitate discovery at the cell-type level, there is a pressing need for computational deconvolution methods capable of leveraging the multitude of underutilized bulk profiles already collected across various organisms, tissues, and conditions. Here, we introduce Unico, a unified cross-omics method designed to deconvolve standard 2-dimensional bulk matrices of samples by features into 3-dimensional tensors representing samples by features by cell types. Unico stands out as the first principled model-based deconvolution method that is theoretically justified for any heterogeneous genomic data. Through the deconvolution of bulk gene expression and DNA methylation datasets, we demonstrate that the transferability of Unico across different data modalities translates into superior performance compared to existing approaches. This advancement enhances our capability to conduct powerful large-scale genomic studies at cell-type resolution without the need for cell sorting or single-cell biology. An R implementation of Unico is available on CRAN.

genomics↗

bioRxiv2022

Microbiome source tracking using single nucleotide variants

Elucidating the sources of a microbiome can provide insight into the ecological dynamics responsible for the formation of these communities. "Source tracking" approaches to date leverage species abundance information, however, single nucleotide variants (SNVs) may be more informative because of their high specificity to certain sources. To overcome the computational burden of utilizing all SNVs for a given sample, we introduce a novel method to identify signature SNVs for source tracking. We show that signature SNVs used as input into a previously designed source tracking algorithm, FEAST, can more accurately estimate contributions than species and provide novel insights, demonstrated in three case studies.

bioinformatics↗

bioRxiv2021

Multi-context genetic modeling of transcriptional regulation resolves novel disease loci

A majority of the variants identified in genome-wide association studies fall in non-coding regions of the genome, indicating their mechanism of impact is mediated via gene expression. Leveraging this hypothesis, transcriptome-wide association studies (TWAS) have assisted in both the interpretation and discovery of additional genes associated with complex traits. However, existing methods for conducting TWAS do not take full advantage of the intra-individual correlation inherently present in multi-context expression studies and do not properly adjust for multiple testing across contexts. We developed CONTENT-- a computationally efficient method with proper cross-context false discovery correction that leverages correlation structure across contexts to improve power and generate context-specific and context-shared components of expression. We applied CONTENT to bulk multi-tissue and single-cell RNA-seq data sets and show that CONTENT leads to a 42% (bulk) and 110% (single cell) increase in the number of genetically predicted genes relative to previous approaches. Interestingly, we find the context-specific component of expression comprises 30% of heritability in tissue-level bulk data and 75% in single-cell data, consistent with cell type heterogeneity in bulk tissue. In the context of TWAS, CONTENT increased the number of gene-phenotype associations discovered by over 47% relative to previous methods across 22 complex traits.

genomics↗

bioRxiv2021

Correcting for Background Noise Improves Phenotype Prediction from Human Gut Microbiome Data

The ability to predict human phenotypes accurately from metagenomic data is crucial for developing biomarkers and therapeutics for diseases. However, metagenomic data is commonly affected by technical or biological variables, unrelated to the phenotype of interest, such as sequencing protocol or host sex, which can greatly reduce or, when correlated to the phenotype of interest, inflate prediction accuracy. We perform a comparative analysis of the ability of different data transformations and existing supervised and unsupervised methods to correct microbiome data for background noise. We find that supervised methods are limited because they cannot account for unmeasured sources of variation. In addition, we observe that unsupervised approaches are often superior in addressing these issues, but existing methods developed for other omic data types, e.g., gene expression and methylation, are restricted by parametric assumptions unsuitable for microbiome data, which is typically compositional, highly skewed, and sparse. We show that application of the centered log-ratio transformation prior to correction with unsupervised approaches improves prediction accuracy for many phenotypes while simultaneously reducing variance due to unwanted sources of variation. As new and larger metagenomic datasets become increasingly available, background noise correction will become essential for generating reproducible microbiome analyses.

bioinformatics↗

bioRxiv2021

Calling differential DNA methylation at cell-type resolution: addressing misconceptions and best practices

We benchmarked two approaches for the detection of cell-type-specific differential DNA methylation: Tensor Composition Analysis (TCA) and a regression model with interaction terms (CellDMC). Our experiments alongside rigorous mathematical explanations show that TCA is superior over CellDMC, thus resolving recent criticisms suggested by Jing et al. Following misconceptions by Jing and colleagues with modelling cell-type-specificity and the application of TCA, we further discuss best practices for performing association studies at cell-type resolution. The scripts for reproducing all of our results and figures are publicly available at github.com/cozygene/CellTypeSpecificMethylationAnalysis.

genomics↗

bioRxiv2020

BATMAN: fast and accurate integration of single-cell RNA-Seq datasets via minimum-weight matching

Single-cell RNA-Sequencing (scRNA-Seq) is a set of technologies used to profile gene expression at the level of individual cells. Although the throughput of scRNA-Seq experiments is steadily growing in terms of the number of cells, large datasets are not yet commonly used due to prohibitively high costs. Integrating multiple datasets into one can improve power in scRNA-Seq experiments, and efficient integration is very important for downstream analyses such as identifying cell-type-specific eQTLs. State-of-the-art scRNA-Seq integration methods are based on the mutual nearest neighbors paradigm and fail to both correct for batch effects and maintain the local structure of the datasets. In this paper, we propose a novel scRNA-Seq dataset integration method called BATMAN (BATch integration via minimum-weight MAtchiNg). Across multiple simulations and real datasets, we show that our method significantly outperforms state-of-the-art tools with respect to existing metrics for batch effects by up to 80% while retaining cell-to-cell relationships. BATMAN is available at https://github.com/mandricigor/batman.

bioinformatics↗

bioRxiv2019

Enhancing droplet-based single-nucleus RNA-seq resolution using the semi-supervised machine learning classifier DIEM

Single-nucleus RNA sequencing (snRNA-seq) measures gene expression in individual nuclei instead of cells, allowing for unbiased cell type characterization in solid tissues. Contrary to single-cell RNA seq (scRNA-seq), we observe that snRNA-seq is commonly subject to contamination by high amounts of extranuclear background RNA, which can lead to identification of spurious cell types in downstream clustering analyses if overlooked. We present a novel approach to remove debris-contaminated droplets in snRNA-seq experiments, called Debris Identification using Expectation Maximization (DIEM). Our likelihood-based approach models the gene expression distribution of debris and cell types, which are estimated using EM. We evaluated DIEM using three snRNA-seq data sets: 1) human differentiating preadipocytes in vitro, 2) fresh mouse brain tissue, and 3) human frozen adipose tissue (AT) from six individuals. All three data sets showed various degrees of extranuclear RNA contamination. We observed that existing methods fail to account for contaminated droplets and led to spurious cell types. When compared to filtering using these state of the art methods, DIEM better removed droplets containing high levels of extranuclear RNA and led to higher quality clusters. Although DIEM was designed for snRNA-seq data, we also successfully applied DIEM to single-cell data. To conclude, our novel method DIEM removes debris-contaminated droplets from single-cell-based data fast and effectively, leading to cleaner downstream analysis. Our code is freely available for use at https://github.com/marcalva/diem.

bioinformatics↗

bioRxiv2019

Optimal design of single-cell RNA sequencing experiments for cell-type-specific eQTL analysis

Single-cell RNA-sequencing (scRNA-Seq) is a compelling approach to simultaneously measure cellular composition and state which is impossible with bulk profiling approaches. However, it has not yet become a widely used tool in population-scale analyses, due to its prohibitively high cost. Here we show that given the same budget, the statistical power of cell-type-specific expression quantitative trait loci (eQTL) mapping can be increased through low-coverage per-cell sequencing of more samples rather than high-coverage sequencing of fewer samples. We also show that multiple experimental designs with different numbers of samples, cells per sample and reads per cell could have similar statistical power, and choosing an appropriate design can yield large cost savings especially when multiplexed workflows are considered. Finally, we provide a practical approach on selecting cost-effective designs for maximizing cell-type-specific eQTL power.

bioinformatics↗

bioRxiv2019

Scalable probabilistic PCA for large-scale genetic variation data

Principal component analysis (PCA) is a key tool for understanding population structure and controlling for population stratification in genome-wide association studies (GWAS). With the advent of large-scale datasets of genetic variation, there is a need for methods that can compute principal components (PCs) with scalable computational and memory requirements. We present ProPCA, a highly scalable method based on a probabilistic generative model, which computes the top PCs on genetic variation data efficiently. We applied ProPCA to compute the top five PCs on genotype data from the UK Biobank, consisting of 488,363 individuals and 146,671 SNPs, in less than thirty minutes. Leveraging the population structure inferred by ProPCA within the White British individuals in the UK Biobank, we scanned for SNPs that are not well-explained by the PCs to identify several novel genome-wide signals of recent putative selection including missense mutations in RPGRIP1L and TLR4.\n\nAuthor SummaryPrincipal component analysis is a commonly used technique for understanding population structure and genetic variation. With the advent of large-scale datasets that contain the genetic information of hundreds of thousands of individuals, there is a need for methods that can compute principal components (PCs) with scalable computational and memory requirements. In this study, we present ProPCA, a highly scalable statistical method to compute genetic PCs efficiently. We systematically evaluate the accuracy and robustness of our method on large-scale simulated data and apply it to the UK Biobank. Leveraging the population structure inferred by ProPCA within the White British individuals in the UK Biobank, we identify several novel signals of putative recent selection.

bioinformatics↗

bioRxiv2019

Accurate estimation of cell composition in bulk expression through robust integration of single-cell information

We present Bisque, a tool for estimating cell type proportions in bulk expression. Bisque implements a regression-based approach that utilizes single-cell RNA-seq (scRNA-seq) data to generate a reference expression profile and learn gene-specific bulk expression transformations to robustly decompose RNA-seq data. These transformations significantly improve decomposition performance compared to existing methods when there is significant technical variation in the generation of the reference profile and observed bulk expression. Importantly, compared to existing methods, our approach is extremely efficient, making it suitable for the analysis of large genomic datasets that are becoming ubiquitous. When applied to subcutaneous adipose and dorsolateral prefrontal cortex expression datasets with both bulk RNA-seq and single-nucleus RNA-seq (snRNA-seq) data, Bisque was able to replicate previously reported associations between cell type proportions and measured phenotypes across abundant and rare cell types. Bisque requires a single-cell reference dataset that reflects physiological cell type composition and can further leverage datasets that includes both bulk and single cell measurements over the same samples for improved accuracy. We further propose an additional mode of operation that merely requires a set of known marker genes. Bisque is available as an R package at: https://github.com/cozygene/bisque.

bioinformatics↗

bioRxiv2019

Distinguishing biological from technical sources of variation by leveraging multiple methylation datasets

DNA methylation remains one of the most widely studied epigenetic markers. One of the major challenges in population studies of methylation is the presence of global methylation effects that may mask local signals. Such global effects may be due to either technical effects (e.g., batch effects) or biological effects (e.g., cell-type composition, genetics). Many methods have been developed for the detection of such global effects, typically in the context of epigenome-wide association studies. However, current unsupervised methods do not distinguish between biological and technical effects, resulting in a loss of highly relevant information. Though supervised methods can be used to estimate known biological effects, it remains difficult to identify and estimate unknown biological effects that globally affect the methylome. Here, we propose CONFINED, a reference-free method based on sparse canonical correlation analysis that captures replicable sources of variation--such as age, sex, and cell-type composition--across multiple methylation datasets and distinguishes them from dataset-specific sources of variability (e.g., technical effects). Consequently, we demonstrate through simulated and real data that by leveraging multiple datasets simultaneously, our approach captures several replicable sources of biological variation better than previous reference-free methods and is considerably more robust to technical noise than previous reference-free methods. CONFINED is available as an R package as detailed at https://github.com/cozygene/CONFINED.

bioinformatics↗

bioRxiv2018

Cell-type-specific resolution epigenetics without the need for cell sorting or single-cell biology

High costs and technical limitations of cell sorting and single-cell techniques currently restrict the collection of large-scale, cell-type-specific DNA methylation data. This, in turn, impedes our ability to tackle key biological questions that pertain to variation within a population, such as identification of disease-associated genes at a cell-type-specific resolution. Here, we show mathematically and empirically that cell-type-specific methylation levels of an individual can be learned from its tissue-level bulk data, conceptually emulating the case where the individual has been profiled with a single-cell resolution and then signals were aggregated in each cell population separately. Provided with this unprecedented way to perform powerful large-scale epigenetic studies with cell-type-specific resolution, we revisit previous studies with tissue-level bulk methylation and reveal novel associations with leukocyte composition in blood and with rheumatoid arthritis. For the latter, we further show consistency with validation data collected from sorted leukocyte sub-types. Corresponding software is available from: https://github.com/cozygene/TCA.

bioinformatics↗

bioRxiv2018

Preoperative predictions of in-hospital mortality using electronic medical record data

BackgroundPredicting preoperative in-hospital mortality using readily-available electronic medical record (EMR) data can aid clinicians in accurately and rapidly determining surgical risk. While previous work has shown that the American Society of Anesthesiologists (ASA) Physical Status Classification is a useful, though subjective, feature for predicting surgical outcomes, obtaining this classification requires a clinician to review the patients medical records. Our goal here is to create an improved risk score using electronic medical records and demonstrate its utility in predicting in-hospital mortality without requiring clinician-derived ASA scores.\n\nMethodsData from 49,513 surgical patients were used to train logistic regression, random forest, and gradient boosted tree classifiers for predicting in-hospital mortality. The features used are readily available before surgery from EMR databases. A gradient boosted tree regression model was trained to impute the ASA Physical Status Classification, and this new, imputed score was included as an additional feature to preoperatively predict in-hospital post-surgical mortality. The preoperative risk prediction was then used as an input feature to a deep neural network (DNN), along with intraoperative features, to predict postoperative in-hospital mortality risk. Performance was measured using the area under the receiver operating characteristic (ROC) curve (AUC).\n\nResultsWe found that the random forest classifier (AUC 0.921, 95%CI 0.908-0.934) outperforms logistic regression (AUC 0.871, 95%CI 0.841-0.900) and gradient boosted trees (AUC 0.897, 95%CI 0.881-0.912) in predicting in-hospital post-surgical mortality. Using logistic regression, the ASA Physical Status Classification score alone had an AUC of 0.865 (95%CI 0.848-0.882). Adding preoperative features to the ASA Physical Status Classification improved the random forest AUC to 0.929 (95%CI 0.915-0.943). Using only automatically obtained preoperative features with no clinician intervention, we found that the random forest model achieved an AUC of 0.921 (95%CI 0.908-0.934). Integrating the preoperative risk prediction into the DNN for postoperative risk prediction results in an AUC of 0.924 (95%CI 0.905-0.941), and with both a preoperative and postoperative risk score for each patient, we were able to show that the mortality risk changes over time.\n\nConclusionsFeatures easily extracted from EMR data can be used to preoperatively predict the risk of in-hospital post-surgical mortality in a fully automated fashion, with accuracy comparable to models trained on features that require clinical expertise. This preoperative risk score can then be compared to the postoperative risk score to show that the risk changes, and therefore should be monitored longitudinally over time.\n\nAuthor summaryRapid, preoperative identification of those patients at highest risk for medical complications is necessary to ensure that limited infrastructure and human resources are directed towards those most likely to benefit. Existing risk scores either lack specificity at the patient level, or utilize the American Society of Anesthesiologists (ASA) physical status classification, which requires a clinician to review the chart. In this manuscript we report on using machine-learning algorithms, specifically random forest, to create a fully automated score that predicts preoperative in-hospital mortality based solely on structured data available at the time of surgery. This score has a higher AUC than both the ASA physical status score and the Charlson comorbidity score. Additionally, we integrate this score with a previously published postoperative score to demonstrate the extent to which patient risk changes during the perioperative period.

bioinformatics↗

bioRxiv2017

CRISPys: Optimal sgRNA design for editing multiple members of a gene family using the CRISPR system

The discovery and development of the CRISPR-Cas9 system in the past few years has made eukaryotic genome editing, and specifically gene knockout for reverse genetics, a simpler, efficient, and effective task. The system is directed to the genomic target site by a programmed single-guide RNA (sgRNA) that base-pairs with the DNA target, subsequently leading to site-specific double-strand breaks. However, many gene families in eukaryotic genomes exhibit partially overlapping functions and, thus, the knockout of one gene might be concealed by the function of the other. In such cases, the reduced specificity of the CRISPR-Cas9 system, which may lead to the cleavage of genomic sites that are not identical to the sgRNA, can be harnessed for the simultaneous knockout of multiple homologous genes. Here, we introduce CRISPys, an algorithm for the optimal design of sgRNAs that would potentially target multiple members of a given gene family. CRISPys first clusters all the potential targets in the input sequences into a hierarchical tree structure that specifies the similarity among them. Then, sgRNAs are proposed in the internal nodes of the tree by embedding mismatches where needed, such that the cleavage efficiencies of the induced targets are maximized. We suggest several approaches for designing the optimal individual sgRNA, and an approach that provides a set of sgRNAs that also accounts for the homologous relationships among gene-family members. We further show by in-silico examination over all gene families in the Solanum lycopersicum genome that our suggested approach outperforms simpler alignment-based techniques.\n\nGraphical abstract O_FIG_DISPLAY_L [Figure 1] M_FIG_DISPLAY C_FIG_DISPLAY\n\nHighlightsO_LIMany genes in eukaryotic genomes exhibit partially overlapping functions. This imposes difficulties on reverse-genetics, as the knockout of one gene might be concealed by the function of the other.\nC_LIO_LIWe present CRISPys, a graph-based algorithm for the optimal design of CRISPR systems given a set of redundant genes.\nC_LIO_LICRISPys harnesses the lack of specificity of the CRISPR-Cas9 genome editing technique, providing researchers the ability to simultaneously mutate multiple genes.\nC_LIO_LIWe show that CRISPys outperforms existing approaches that are based on simple alignment of the input gene family.\nC_LI

bioinformatics↗

bioRxiv2017

Modeling the temporal dynamics of the gut microbial community in adults and infants

Given the highly dynamic and complex nature of the human gut microbial community, the ability to identify and predict time-dependent compositional patterns of microbes is crucial to our understanding of the structure and function of this ecosystem. One factor that could affect such time-dependent patterns is microbial interactions, wherein community composition at a given time point affects the microbial composition at a later time point. However, the field has not yet settled on the degree of this effect. Specifically, it has been recently suggested that only a minority of the operational taxonomic units (OTUs) depend on the microbial composition in earlier times. To address the issue of identifying and predicting temporal microbial patterns we developed a new model, MTV-LMM (Microbial Temporal Variability Linear Mixed Model), a linear mixed model for the prediction of the microbial community temporal dynamics based on the community composition at previous time stamps. MTV-LMM can identify time-dependent microbes in time series datasets, which can then be used to analyze the trajectory of the microbiome over time. We evaluated the performance of MTV-LMM on three human microbiome time series datasets, and found that MTV-LMM significantly outperforms all existing methods for microbiome time series modeling. Particularly, we demonstrate that the effect of the microbial composition in previous time points on the abundance levels of an OTU at a later time point is underestimated by a factor of at least 10 when applying previous approaches. Using MTV-LMM, we demonstrate that a considerable proportion of the human gut microbiome, both in infants and adults, has a significant time-dependent component that can be predicted based on microbiome composition in earlier time points. This suggests that microbiome composition at a given time point is a major factor in defining future microbiome composition and that this phenomenon is considerably more common than previously reported for the human gut microbiome.

microbiology↗

bioRxiv2017

An exact and efficient score test for variance components models

Testing for the existence of variance components in linear mixed models is a fundamental task in many applicative fields. In statistical genetics, the score test has recently become instrumental in the task of testing an association between a set of genetic markers and a phenotype. With few markers, this amounts to set-based variance component tests, which attempt to increase power in association studies by aggregating weak individual effects. When the entire genome is considered, it allows testing for the heritability of a phenotype, defined as the proportion of phenotypic variance explained by genetics. In the popular score-based Sequence Kernel Association Test (SKAT) method, the assumed distribution of the score test statistic is uncalibrated in small samples, with a correction being computationally expensive. This may cause severe inflation or deflation of p-values, even when the null hypothesis is true. Here, we characterize the conditions under which this discrepancy holds, and show it may occur also in large real datasets, such as a dataset from the Wellcome Trust Case Control Consortium 2 (n=13,950) study, and in particular when the individuals in the sample are unrelated. In these cases the SKAT approximation tends to be highly over-conservative and therefore underpowered. To address this limitation, we suggest an efficient method to calculate exact p-values for the score test in the case of a single variance component and a continuous response vector, which can speed up the analysis by orders of magnitude. Our results enable fast and accurate application of the score test in heritability and in set-based association tests. Our method is available in http://github.com/cozygene/RL-SKAT.

genetics↗