Mapping disease regulatory circuits at cell-type resolution from single-cell multiomics data.

Resolving chromatin-remodeling-linked gene expression changes at cell-type resolution is important for understanding disease states. Here we describe MAGICAL (Multiome Accessibility Gene Integration Calling and Looping), a hierarchical Bayesian approach that leverages paired single-cell RNA sequencing and single-cell transposase-accessible chromatin sequencing from different conditions to map disease-associated transcription factors, chromatin sites, and genes as regulatory circuits. By simultaneously modeling signal variation across cells and conditions in both omics data types, MAGICAL achieved high accuracy on circuit inference. We applied MAGICAL to study Staphylococcus aureus sepsis from peripheral blood mononuclear single-cell data that we generated from subjects with bloodstream infection and uninfected controls. MAGICAL identified sepsis-associated regulatory circuits predominantly in CD14 monocytes, known to be activated by bacterial sepsis. We addressed the challenging problem of distinguishing host regulatory circuit responses to methicillin-resistant and methicillin-susceptible S. aureus infections. Although differential expression analysis failed to show predictive value, MAGICAL identified epigenetic circuit biomarkers that distinguished methicillin-resistant from methicillin-susceptible S. aureus infections.

Atlas of primary cell-type-specific sequence models of gene expression and variant effects.

Human biology is rooted in highly specialized cell types programmed by a common genome, 98% of which is outside of genes. Genetic variation in the enormous noncoding space is linked to the majority of disease risk. To address the problem of linking these variants to expression changes in primary human cells, we introduce ExPectoSC, an atlas of modular deep-learning-based models for predicting cell-type-specific gene expression directly from sequence. We provide models for 105 primary human cell types covering 7 organ systems, demonstrate their accuracy, and then apply them to prioritize relevant cell types for complex human diseases. The resulting atlas of sequence-based gene expression and variant effects is publicly available in a user-friendly interface and readily extensible to any primary cell types. We demonstrate the accuracy of our approach through systematic evaluations and apply the models to prioritize ClinVar clinical variants of uncertain significance, verifying our top predictions experimentally.

Decoding disease: from genomes to networks to phenotypes.

Interpreting the effects of genetic variants is key to understanding individual susceptibility to disease and designing personalized therapeutic approaches. Modern experimental technologies are enabling the generation of massive compendia of human genome sequence data and associated molecular and phenotypic traits, together with genome-scale expression, epigenomics and other functional genomic data. Integrative computational models can leverage these data to understand variant impact, elucidate the effect of dysregulated genes on biological pathways in specific disease and tissue contexts, and interpret disease risk beyond what is feasible with experiments alone. In this Review, we discuss recent developments in machine learning algorithms for genome interpretation and for integrative molecular-level modelling of cells, tissues and organs relevant to disease. More specifically, we highlight existing methods and key challenges and opportunities in identifying specific disease-causing genetic variants and linking them to molecular pathways and, ultimately, to disease phenotypes.

Modeling transcriptional regulation of model species with deep learning.

To enable large-scale analyses of transcription regulation in model species, we developed DeepArk, a set of deep learning models of the cis-regulatory activities for four widely studied species: Caenorhabditis elegans, Danio rerio, Drosophila melanogaster, and Mus musculus DeepArk accurately predicts the presence of thousands of different context-specific regulatory features, including chromatin states, histone marks, and transcription factors. In vivo studies show that DeepArk can predict the regulatory impact of any genomic variant (including rare or not previously observed) and enables the regulatory annotation of understudied model species.

Tissue-specific enhancer functional networks for associating distal regulatory regions to disease.

Systematic study of tissue-specific function of enhancers and their disease associations is a major challenge. We present an integrative machine-learning framework, FENRIR, that integrates thousands of disparate epigenetic and functional genomics datasets to infer tissue-specific functional relationships between enhancers for 140 diverse human tissues and cell types, providing a regulatory-region-centric approach to systematically identify disease-associated enhancers. We demonstrated its power to accurately prioritize enhancers associated with 25 complex diseases. In a case study on autism, FENRIR-prioritized enhancers showed a significant proband-specific de novo mutation enrichment in a large, sibling-controlled cohort, indicating pathogenic signal. We experimentally validated transcriptional regulatory activities of eight enhancers, including enhancers not previously reported with autism, and demonstrated their differential regulatory potential between proband and sibling alleles. Thus, FENRIR is an accurate and effective framework for the study of tissue-specific enhancers and their role in disease. FENRIR can be accessed at fenrir.flatironinstitute.org/.

Genome-wide landscape of RNA-binding protein target site dysregulation reveals a major impact on psychiatric disorder risk.

Despite the strong genetic basis of psychiatric disorders, the underlying molecular mechanisms are largely unmapped. RNA-binding proteins (RBPs) are responsible for most post-transcriptional regulation, from splicing to translation to localization. RBPs thus act as key gatekeepers of cellular homeostasis, especially in the brain. However, quantifying the pathogenic contribution of noncoding variants impacting RBP target sites is challenging. Here, we leverage a deep learning approach that can accurately predict the RBP target site dysregulation effects of mutations and discover that RBP dysregulation is a principal contributor to psychiatric disorder risk. RBP dysregulation explains a substantial amount of heritability not captured by large-scale molecular quantitative trait loci studies and has a stronger impact than common coding region variants. We share the genome-wide profiles of RBP dysregulation, which we use to identify DDHD2 as a candidate schizophrenia risk gene. This resource provides a new analytical framework to connect the full range of RNA regulation to complex disease.

Modeling molecular development of breast cancer in canine mammary tumors.

Understanding the changes in diverse molecular pathways underlying the development of breast tumors is critical for improving diagnosis, treatment, and drug development. Here, we used RNA-profiling of canine mammary tumors (CMTs) coupled with a robust analysis framework to model molecular changes in human breast cancer. Our study leveraged a key advantage of the canine model, the frequent presence of multiple naturally occurring tumors at diagnosis, thus providing samples spanning normal tissue and benign and malignant tumors from each patient. We showed human breast cancer signals, at both expression and mutation level, are evident in CMTs. Profiling multiple tumors per patient enabled by the CMT model allowed us to resolve statistically robust transcription patterns and biological pathways specific to malignant tumors versus those arising in benign tumors or shared with normal tissues. We showed that multiple histological samples per patient is necessary to effectively capture these progression-related signatures, and that carcinoma-specific signatures are predictive of survival for human breast cancer patients. To catalyze and support similar analyses and use of the CMT model by other biomedical researchers, we provide FREYA, a robust data processing pipeline and statistical analyses framework.

Genomic RNA Elements Drive Phase Separation of the SARS-CoV-2 Nucleocapsid.

We report that the SARS-CoV-2 nucleocapsid protein (N-protein) undergoes liquid-liquid phase separation (LLPS) with viral RNA. N-protein condenses with specific RNA genomic elements under physiological buffer conditions and condensation is enhanced at human body temperatures (33°C and 37°C) and reduced at room temperature (22°C). RNA sequence and structure in specific genomic regions regulate N-protein condensation while other genomic regions promote condensate dissolution, potentially preventing aggregation of the large genome. At low concentrations, N-protein preferentially crosslinks to specific regions characterized by single-stranded RNA flanked by structured elements and these features specify the location, number, and strength of N-protein binding sites (valency). Liquid-like N-protein condensates form in mammalian cells in a concentration-dependent manner and can be altered by small molecules. Condensation of N-protein is RNA sequence and structure specific, sensitive to human body temperature, and manipulatable with small molecules, and therefore presents a screenable process for identifying antiviral compounds effective against SARS-CoV-2.

SARS-CoV-2 receptor networks in diabetic and COVID-19-associated kidney disease.

COVID-19 morbidity and mortality are increased via unknown mechanisms in patients with diabetes and kidney disease. SARS-CoV-2 uses angiotensin-converting enzyme 2 (ACE2) for entry into host cells. Because ACE2 is a susceptibility factor for infection, we investigated how diabetic kidney disease and medications alter ACE2 receptor expression in kidneys. Single cell RNA profiling of kidney biopsies from healthy living donors and patients with diabetic kidney disease revealed ACE2 expression primarily in proximal tubular epithelial cells. This cell-specific localization was confirmed by in situ hybridization. ACE2 expression levels were unaltered by exposures to renin-angiotensin-aldosterone system inhibitors in diabetic kidney disease. Bayesian integrative analysis of a large compendium of public -omics datasets identified molecular network modules induced in ACE2-expressing proximal tubular epithelial cells in diabetic kidney disease (searchable at hb.flatironinstitute.org/covid-kidney) that were linked to viral entry, immune activation, endomembrane reorganization, and RNA processing. The diabetic kidney disease ACE2-positive proximal tubular epithelial cell module overlapped with expression patterns seen in SARS-CoV-2-infected cells. Similar cellular programs were seen in ACE2-positive proximal tubular epithelial cells obtained from urine samples of 13 hospitalized patients with COVID-19, suggesting a consistent ACE2-coregulated proximal tubular epithelial cell expression program that may interact with the SARS-CoV-2 infection processes. Thus SARS-CoV-2 receptor networks can seed further research into risk stratification and therapeutic strategies for COVID-19-related kidney damage.

SARS-CoV-2 receptor networks in diabetic and COVID-19 associated kidney disease.

COVID-19 morbidity and mortality is increased in patients with diabetes and kidney disease via unknown mechanisms. SARS-CoV-2 uses angiotensin-converting enzyme 2 (ACE2) for entry into host cells. Since ACE2 is a susceptibility factor for infection, we investigated how diabetic kidney disease (DKD) and medications alter ACE2 receptor expression in kidneys. Single cell RNA profiling of healthy living donor (LD) and DKD kidney biopsies revealed ACE2 expression primarily in proximal tubular epithelial cells (PTEC). This cell specific localization was confirmed by in situ hybridization. ACE2 expression levels were unaltered by exposures to renin angiotensin aldosterone system inhibitors in DKD. Bayesian integrative analysis of a large compendium of public -omics datasets identified molecular network modules induced in ACE2-expressing PTEC in DKD (searchable at hb.flatironinstitute.org/covid-kidney) that were linked to viral entry, immune activation, endomembrane reorganization, and RNA processing. The DKD ACE2-positive PTEC module overlapped with expression patterns seen in SARS-CoV-2 infected cells. Similar cellular programs were seen in ACE2-positive PTEC obtained from urine samples of 13 COVID-19 patients who were hospitalized, suggesting a consistent ACE2-coregulated PTEC expression program that may interact with the SARS-CoV-2 infection processes. Thus SARS-CoV-2 receptor networks can seed further research into risk stratification and therapeutic strategies for COVID-19 related kidney damage.

Accurate genome-wide predictions of spatio-temporal gene expression during embryonic development.

Comprehensive information on the timing and location of gene expression is fundamental to our understanding of embryonic development and tissue formation. While high-throughput in situ hybridization projects provide invaluable information about developmental gene expression patterns for model organisms like Drosophila, the output of these experiments is primarily qualitative, and a high proportion of protein coding genes and most non-coding genes lack any annotation. Accurate data-centric predictions of spatio-temporal gene expression will therefore complement current in situ hybridization efforts. Here, we applied a machine learning approach by training models on all public gene expression and chromatin data, even from whole-organism experiments, to provide genome-wide, quantitative spatio-temporal predictions for all genes. We developed structured in silico nano-dissection, a computational approach that predicts gene expression in >200 tissue-developmental stages. The algorithm integrates expression signals from a compendium of 6,378 genome-wide expression and chromatin profiling experiments in a cell lineage-aware fashion. We systematically evaluated our performance via cross-validation and experimentally confirmed 22 new predictions for four different embryonic tissues. The model also predicts complex, multi-tissue expression and developmental regulation with high accuracy. We further show the potential of applying these genome-wide predictions to extract tissue specificity signals from non-tissue-dissected experiments, and to prioritize tissues and stages for disease modeling. This resource, together with the exploratory tools are freely available at our webserver http://find.princeton.edu, which provides a valuable tool for a range of applications, from predicting spatio-temporal expression patterns to recognizing tissue signatures from differential gene expression profiles.

Whole-genome deep-learning analysis identifies contribution of noncoding mutations to autism risk.

We address the challenge of detecting the contribution of noncoding mutations to disease with a deep-learning-based framework that predicts the specific regulatory effects and the deleterious impact of genetic variants. Applying this framework to 1,790 autism spectrum disorder (ASD) simplex families reveals a role in disease for noncoding mutations-ASD probands harbor both transcriptional- and post-transcriptional-regulation-disrupting de novo mutations of significantly higher functional impact than those in unaffected siblings. Further analysis suggests involvement of noncoding mutations in synaptic transmission and neuronal development and, taken together with previous studies, reveals a convergent genetic landscape of coding and noncoding mutations in ASD. We demonstrate that sequences carrying prioritized mutations identified in probands possess allele-specific regulatory activity, and we highlight a link between noncoding mutations and heterogeneity in the IQ of ASD probands. Our predictive genomics framework illuminates the role of noncoding mutations in ASD and prioritizes mutations with high impact for further study, and is broadly applicable to complex human diseases.

A Computational Framework for Genome-wide Characterization of the Human Disease Landscape.

A key challenge for the diagnosis and treatment of complex human diseases is identifying their molecular basis. Here, we developed a unified computational framework, URSAHD (Unveiling RNA Sample Annotation for Human Diseases), that leverages machine learning and the hierarchy of anatomical relationships present among diseases to integrate thousands of clinical gene expression profiles and identify molecular characteristics specific to each of the hundreds of complex diseases. URSAHD can distinguish between closely related diseases more accurately than literature-validated genes or traditional differential-expression-based computational approaches and is applicable to any disease, including rare and understudied ones. We demonstrate the utility of URSAHD in classifying related nervous system cancers and experimentally verifying novel neuroblastoma-associated genes identified by URSAHD. We highlight the applications for potential targeted drug-repurposing and for quantitatively assessing the molecular response to clinical therapies. URSAHD is freely available for public use, including the use of underlying models, at ursahd.princeton.edu.

Deep learning sequence-based ab initio prediction of variant effects on expression and disease risk.

Key challenges for human genetics, precision medicine and evolutionary biology include deciphering the regulatory code of gene expression and understanding the transcriptional effects of genome variation. However, this is extremely difficult because of the enormous scale of the noncoding mutation space. We developed a deep learning-based framework, ExPecto, that can accurately predict, ab initio from a DNA sequence, the tissue-specific transcriptional effects of mutations, including those that are rare or that have not been observed. We prioritized causal variants within disease- or trait-associated loci from all publicly available genome-wide association studies and experimentally validated predictions for four immune-related diseases. By exploiting the scalability of ExPecto, we characterized the regulatory mutation space for human RNA polymerase II-transcribed genes by in silico saturation mutagenesis and profiled > 140 million promoter-proximal mutations. This enables probing of evolutionary constraints on gene expression and ab initio prediction of mutation disease effects, making ExPecto an end-to-end computational framework for the in silico prediction of expression and disease risk.

Genome-wide prediction and functional characterization of the genetic basis of autism spectrum disorder.

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder with a strong genetic basis. Yet, only a small fraction of potentially causal genes-about 65 genes out of an estimated several hundred-are known with strong genetic evidence from sequencing studies. We developed a complementary machine-learning approach based on a human brain-specific gene network to present a genome-wide prediction of autism risk genes, including hundreds of candidates for which there is minimal or no prior genetic evidence. Our approach was validated in a large independent case-control sequencing study. Leveraging these genome-wide predictions and the brain-specific network, we demonstrated that the large set of ASD genes converges on a smaller number of key pathways and developmental stages of the brain. Finally, we identified likely pathogenic genes within frequent autism-associated copy-number variants and proposed genes and pathways that are likely mediators of ASD across multiple copy-number variants. All predictions and functional insights are available at http://asd.princeton.edu.

An extended set of yeast-based functional assays accurately identifies human disease mutations.

We can now routinely identify coding variants within individual human genomes. A pressing challenge is to determine which variants disrupt the function of disease-associated genes. Both experimental and computational methods exist to predict pathogenicity of human genetic variation. However, a systematic performance comparison between them has been lacking. Therefore, we developed and exploited a panel of 26 yeast-based functional complementation assays to measure the impact of 179 variants (101 disease- and 78 non-disease-associated variants) from 22 human disease genes. Using the resulting reference standard, we show that experimental functional assays in a 1-billion-year diverged model organism can identify pathogenic alleles with significantly higher precision and specificity than current computational methods.

Use of the BioGRID Database for Analysis of Yeast Protein and Genetic Interactions.

The BioGRID database is an extensive repository of curated genetic and protein interactions for the budding yeast Saccharomyces cerevisiae, the fission yeast Schizosaccharomyces pombe, and the yeast Candida albicans SC5314, as well as for several other model organisms and humans. This protocol describes how to use the BioGRID website to query genetic or protein interactions for any gene of interest, how to visualize the associated interactions using an embedded interactive network viewer, and how to download data files for either selected interactions or the entire BioGRID interaction data set.

BioGRID: A Resource for Studying Biological Interactions in Yeast.

The Biological General Repository for Interaction Datasets (BioGRID) is a freely available public database that provides the biological and biomedical research communities with curated protein and genetic interaction data. Structured experimental evidence codes, an intuitive search interface, and visualization tools enable the discovery of individual gene, protein, or biological network function. BioGRID houses interaction data for the major model organism species--including yeast, nematode, fly, zebrafish, mouse, and human--with particular emphasis on the budding yeast Saccharomyces cerevisiae and the fission yeast Schizosaccharomyces pombe as pioneer eukaryotic models for network biology. BioGRID has achieved comprehensive curation coverage of the entire literature for these two major yeast models, which is actively maintained through monthly curation updates. As of September 2015, BioGRID houses approximately 335,400 biological interactions for budding yeast and approximately 67,800 interactions for fission yeast. BioGRID also supports an integrated posttranslational modification (PTM) viewer that incorporates more than 20,100 yeast phosphorylation sites curated through its sister database, the PhosphoGRID.