Mapping enhancers and their target genes in specific cell types is crucial for understanding gene regulation and human disease genetics. However, accurately predicting enhancer-gene regulatory interactions from single-cell datasets has been challenging. Here, we introduce a new family of classification models, scE2G, to predict enhancer-gene regulation. These models use features from single-cell ATAC-seq or multiomic RNA and ATAC-seq data and are trained on a CRISPR perturbation dataset including >10,000 evaluated element-gene pairs. We benchmark scE2G models against CRISPR perturbations, fine-mapped eQTLs, and GWAS variant-gene associations and demonstrate state-of-the-art performance at prediction tasks across multiple cell types and categories of perturbations. We apply scE2G to build maps of enhancer-gene regulatory interactions in heterogeneous tissues and interpret noncoding variants associated with complex traits, nominating regulatory interactions linking INPP4B and IL15 to lymphocyte counts. The scE2G models will enable accurate mapping of enhancer-gene regulatory interactions across thousands of diverse human cell types.

Despite a considerable expenditure of time and resources and significant advances in experimental models of disease, cancer research continues to suffer from extremely low success rates in translating preclinical discoveries into clinical practice. The continued failure of cancer drug development, particularly late in the course of human testing, not only impacts patient outcomes, but also drives up the cost for those therapies that do succeed. It is clear that a paradigm shift is necessary if improvements in this process are to occur. One promising direction for increasing translational success is comparative oncology—the study of cancer across species, often involving veterinary patients that develop naturally-occurring cancers. Comparative oncology leverages the power of cross-species analyses to understand the fundamental drivers of cancer protective mechanisms, as well as factors contributing to cancer initiation and progression. Clinical trials in veterinary patients with cancer provide an opportunity to evaluate novel therapeutics in a setting that recapitulates many of the key features of human cancers, including genomic aberrations that underly tumor development, response and resistance to treatment, and the presence of comorbidities that can affect outcomes. With a concerted effort from basic scientists, human physicians and veterinarians, comparative oncology has the potential to enhance the cost-effectiveness and efficiency of pipelines for cancer drug discovery and other cancer treatments.

Background: Osteosarcoma is a rare but aggressive bone cancer that occurs primarily in children. Like other rare cancers, treatment advances for osteosarcoma have stagnated, with little improvement in survival for the past several decades. Developing new treatments has been hampered by extensive genomic heterogeneity and limited access to patient samples to study the biology of this complex disease. Methods: To overcome these barriers, we combined the power of comparative oncology with patient-derived models of cancer and high-throughput chemical screens in a cross-species drug discovery pipeline. Results: Coupling in vitro high-throughput drug screens on low-passage and established cell lines with in vivo validation in patient-derived xenografts we identify the proteasome and CRM1 nuclear export pathways as therapeutic sensitivities in osteosarcoma, with dual inhibition of these pathways inducing synergistic cytotoxicity. Conclusions: These collective efforts provide an experimental framework and set of new tools for osteosarcoma and other rare cancers to identify and study new therapeutic vulnerabilities.

The evolution of therapeutic resistance is a major cause of death for cancer patients. The development of therapy resistance is shaped by the ecological dynamics within the tumor microenvironment and the selective pressure of the host immune system. These selective forces often lead to evolutionary convergence on pathways or hallmarks that drive progression. Thus, a deeper understanding of the evolutionary convergences that occur could reveal vulnerabilities to treat therapy-resistant cancer. To this end, we combined phylogenetic clustering, systems biology analyses, and molecular experimentation to identify convergences in gene expression data onto common signaling pathways. We applied these methods to derive new insights about the networks at play during transforming growth factor-β (TGF-β)-mediated epithelial–mesenchymal transition in lung cancer. Phylogenetic analyses of gene expression data from TGF-β-treated cells revealed convergence of cells toward amine metabolic pathways and autophagy during TGF-β treatment. Knockdown of the autophagy regulatory, ATG16L1, re-sensitized lung cancer cells to cancer therapies following TGF-β-induced resistance, implicating autophagy as a TGF-β-mediated chemoresistance mechanism. In addition, high ATG16L expression was found to be a poor prognostic marker in multiple cancer types. These analyses reveal the usefulness of combining evolutionary and systems biology methods with experimental validation to illuminate new therapeutic vulnerabilities for cancer.

Congenital heart defects (CHD) arise in part due to inherited genetic variants that alter genes and noncoding regulatory elements in the human genome. These variants are thought to act during fetal development to influence the formation of different heart structures. However, identifying the genes, pathways, and cell types that mediate these effects has been challenging due to the diversity of cell types involved in heart development as well as the superimposed complexities of interpreting noncoding sequences. As such, understanding the molecular functions of both noncoding and coding variants remains paramount to our fundamental understanding of cardiac development and CHD. Here, we created a gene regulation map of the healthy human fetal heart across developmental time and applied it to interpret the functions of variants associated with CHD and quantitative cardiac traits. We collected multiomic data from 734,078 single cells sampled from 41 fetal hearts spanning post-conception weeks (PCW) 6 to 22, enabling the construction of gene regulation maps in 90 cardiac cell types and states, including rare populations of cardiac conduction system (CCS) cell types. We find that many such risk variants, both coding and noncoding, map onto gene regulatory networks in valve cells. In an unbiased analysis of 90 cardiac cell types and states, valvular interstitial cells (VICs) show the strongest enrichments for known CHD genes. Nine CHD genes are linked to common noncoding variants associated with other valve diseases or traits via transcriptional enhancers in VICs. CRISPRi experiments show that multiple CHD genes have cell-autonomous effects in VICs, including regulating the expression of other CHD genes and genes involved in cell migration and developmental patterning. Finally, we characterize a CHD-associated noncoding variant whose activity is specific to discrete structures within the heart, illuminating how such variants can impact specific aspects of heart structure and function. Together, these results suggest that genetic risk for many subtypes of CHD is influenced in part by genetic regulation of valvular cells, expanding our understanding of the cell types responsible for CHD beyond the working cardiomyocyte. This regulatory map of the human fetal heart provides a foundational resource for understanding cardiac development, interpreting genetic variants associated with heart disease, and discovering targets for cell-type specific therapies.

A major challenge in human genetics is to identify all distal regulatory elements and determine their effects on target gene expression in a given cell type. To this end, large-scale CRISPR screens have been conducted to perturb thousands of candidate enhancers. Using these data, predictive models have been developed that aim to generalize such findings to predict which enhancers regulate which genes across the genome. However, existing CRISPR methods and large-scale datasets have limitations in power, scale, or selection bias, with the potential to skew our understanding of the properties of distal regulatory elements and confound our ability to evaluate predictive models. Here, we develop a new framework for highly powered, unbiased CRISPR screens, including an optimized experimental method (Direct-Capture Targeted Perturb-seq (DC-TAP-seq)), a random design strategy, and a comprehensive analytical pipeline that accounts for statistical power. We applied this framework to survey 1,425 randomly selected candidate regulatory elements across two human cell lines. Our results reveal fundamental properties of distal regulatory elements in the human genome. Most element-gene regulatory interactions are estimated to have small effect sizes (<10%), which previous experiments were not powered to detect. Most cis -regulatory interactions occur over short genomic distances (<100 kb). A large fraction of the discovered regulatory elements bind CTCF but do not show chromatin marks typical of classical enhancers. Housekeeping genes have similar frequencies of distal regulatory elements compared to other genes, but with 2-fold weaker effect sizes. Comparisons to the predictions of the ENCODE-rE2G model suggest that, while performance is similar across two cell types, new models will be needed to detect elements with weaker effect sizes, regulatory effects of CTCF sites, and enhancers for housekeeping genes. Overall, this study describes the first unbiased, perturbation-based survey of thousands of distal regulatory element-gene connections, and provides a framework for expanding such efforts to build more complete maps of distal regulation in the human genome.