Accurate inference of population structure is important in many studies of population genetics. Here we present HaploNet, a method for performing dimensionality reduction and clustering of genetic data. The method is based on local clustering of phased haplotypes using neural networks from whole-genome sequencing or dense genotype data. By using Gaussian mixtures in a variational autoencoder framework, we are able to learn a low-dimensional latent space in which we cluster haplotypes along the genome in a highly scalable manner. We show that we can use haplotype clusters in the latent space to infer global population structure using haplotype information by exploiting the generative properties of our framework. Based on fitted neural networks and their latent haplotype clusters, we can perform principal component analysis and estimate ancestry proportions based on a maximum likelihood framework. Using sequencing data from simulations and closely related human populations, we show that our approach is better at distinguishing closely related populations than standard admixture and principal component analysis software. We further show that HaploNet is fast and highly scalable by applying it to genotype array data of the UK Biobank.

We present fastGLOBETROTTER, an efficient new haplotype-based technique to identify, date, and describe admixture events using genome-wide autosomal data. With simulations, we show how fastGLOBETROTTER reduces computation time by an order of magnitude relative to the related technique GLOBETROTTER without suffering loss of accuracy. We apply fastGLOBETROTTER to a cohort of more than 6000 Europeans from 10 countries, revealing previously unreported admixture signals. In particular, we infer multiple periods of admixture related to East Asian or Siberian-like sources, starting >2000 yr ago, in people living in countries north of the Baltic Sea. In contrast, we infer admixture related to West Asian, North African, and/or Southern European sources in populations south of the Baltic Sea, including admixture dated to ∼300-700 CE, overlapping the fall of the Roman Empire, in people from Belgium, France, and parts of Germany. Our new approach scales to analyzing hundreds to thousands of individuals from a putatively admixed population and, hence, is applicable to emerging large-scale cohorts of genetically homogeneous populations.

Analysis of allele-specific gene expression (ASE) is a powerful approach for studying gene regulation, particularly when sample sizes are small, such as for rare diseases, or when studying the effects of rare genetic variation. However, detection of ASE events relies on accurate alignment of RNA sequencing reads, where challenges still remain, particularly for reads containing genetic variants or those that align to many different genomic locations. We have developed the Personalised ASE Caller (PAC), a tool that combines multiple steps to improve the quantification of allelic reads, including personalized (i.e., diploid) read alignment with improved allocation of multimapping reads. Using simulated RNA sequencing data, we show that PAC outperforms standard alignment approaches for ASE detection, reducing the number of sites with incorrect biases (>10%) by ∼80% and increasing the number of sites that can be reliably quantified by ∼3%. Applying PAC to real RNA sequencing data from 670 whole-blood samples, we show that genetic regulatory signatures inferred from ASE data more closely match those from population-based methods that are less prone to alignment biases. Finally, we use PAC to characterize cell type-specific ASE events that would be missed by standard alignment approaches, and in doing so identify disease relevant genes that may modulate their effects through the regulation of gene expression. PAC can be applied to the vast quantity of existing RNA sequencing data sets to better understand a wide array of fundamental biological and disease processes.

Recent assemblies by the T2T and VGP consortia have achieved significant accuracy but required a tremendous amount of effort and resources. More typical assembly efforts, on the other hand, still suffer both from misassemblies (joining sequences that should not be adjacent) and from underassemblies (not joining sequences that should be adjacent). To better understand the common algorithm-driven causes of these limitations, we investigated the unitig algorithm, which is a core algorithm at the heart of most assemblers. We prove that, contrary to popular belief, even when there are no sequencing errors, unitigs are not always safe (i.e., they are not guaranteed to be substrings of the sequenced genome). We also prove that the unitigs of a bidirected de Bruijn graph are different from those of a doubled de Bruijn graph and, contrary to our expectations, result in underassembly. Using experimental simulations, we then confirm that these two artifacts exist not only in theory but also in the output of widely used assemblers. In particular, when coverage is low, then even error-free data result in unsafe unitigs; also, unitigs may unnecessarily split palindromes in half if special care is not taken. To the best of our knowledge, this paper is the first to theoretically predict the existence of these assembler artifacts and confirm and measure the extent of their occurrence in practice.

Histone modifications are critical epigenetic indicators of chromatin state associated with gene expression. Although the reprogramming patterns of H3K4me3 and H3K27me3 have been elucidated in mouse and human preimplantation embryos, the relationship between these marks and zygotic genome activation (ZGA) remains poorly understood. By ultra-low-input native chromatin immunoprecipitation and sequencing, we profiled global H3K4me3 and H3K27me3 in porcine oocytes and in vitro fertilized (IVF) embryos. We found that promoters of ZGA genes occupied sharp H3K4me3 peaks in oocytes, and these peaks became broader after fertilization, and reshaped into sharp again during ZGA. By simultaneous depletion of H3K4me3 demethylase KDM5B and KDM5C, we determined that broad H3K4me3 domain maintenance impaired ZGA gene expression, suggesting its function to prevent premature ZGA entry. By contrast, broad H3K27me3 domains underwent global removal upon fertilization, followed by a re-establishment for H3K4me3/H3K27me3 bivalency in morulae. We also found that bivalent marks were deposited at promoters of ZGA genes, and inhibiting this deposition was correlated with the activation of ZGA genes. It suggests that promoter bivalency contributes to ZGA exit in porcine embryos. Moreover, we demonstrated that aberrant reprogramming of H3K4me3 and H3K27me3 triggered ZGA dysregulation in somatic cell nuclear transfer (SCNT) embryos, whereas H3K27me3-mediated imprinting did not exist in porcine IVF and SCNT embryos. Our findings highlight two previously unknown epigenetic reprogramming modes coordinated with ZGA in porcine preimplantation embryos. Finally, the similarities observed between porcine and human histone modification dynamics suggest that the porcine embryo may also be a useful model for human embryo research.

Genomic rearrangements are known to result in proto-oncogene deregulation in many cancers, but the link to 3D genome structure remains poorly understood. Here, we used the highly predictive heteromorphic polymer (HiP-HoP) model to predict chromatin conformations at the proto-oncogene CCND1 in healthy and malignant B cells. After confirming that the model gives good predictions of Hi-C data for the nonmalignant human B cell-derived cell line GM12878, we generated predictions for two cancer cell lines, U266 and Z-138. These possess genome rearrangements involving CCND1 and the immunoglobulin heavy locus (IGH), which we mapped using targeted genome sequencing. Our simulations showed that a rearrangement in U266 cells where a single IGH super-enhancer is inserted next to CCND1 leaves the local topologically associated domain (TAD) structure intact. We also observed extensive changes in enhancer-promoter interactions within the TAD, suggesting that it is the downstream chromatin remodeling which gives rise to the oncogene activation, rather than the presence of the inserted super-enhancer DNA sequence per se. Simulations of the IGH-CCND1 reciprocal translocation in Z-138 cells revealed that an oncogenic fusion TAD is created, encompassing CCND1 and the IGH super-enhancers. We predicted how the structure and expression of CCND1 changes in these different cell lines, validating this using qPCR and fluorescence in situ hybridization microscopy. Our work demonstrates the power of polymer simulations to predict differences in chromatin interactions and gene expression for different translocation breakpoints.

Assembling RNA-seq reads into full-length transcripts is crucial in transcriptomic studies and poses computational challenges. Here we present TransMeta, a simple and robust algorithm that simultaneously assembles RNA-seq reads from multiple samples. TransMeta is designed based on the newly introduced vector-weighted splicing graph model, which enables accurate reconstruction of the consensus transcriptome via incorporating a cosine similarity-based combing strategy and a newly designed label-setting path-searching strategy. Tests on both simulated and real data sets show that TransMeta consistently outperforms PsiCLASS, StringTie2 plus its merge mode, and Scallop plus TACO, the most popular tools, in terms of precision and recall under a wide range of coverage thresholds at the meta-assembly level. Additionally, TransMeta consistently shows superior performance at the individual sample level.

Over the past 20 years, the explosion of genomic data collection and the cloud computing revolution have made computational and data science research accessible to anyone with a web browser and an internet connection. However, students at institutions with limited resources have received relatively little exposure to curricula or professional development opportunities that lead to careers in genomic data science. To broaden participation in genomics research, the scientific community needs to support these programs in local education and research at underserved institutions (UIs). These include community colleges, historically Black colleges and universities, Hispanic-serving institutions, and tribal colleges and universities that support ethnically, racially, and socioeconomically underrepresented students in the United States. We have formed the Genomic Data Science Community Network to support students, faculty, and their networks to identify opportunities and broaden access to genomic data science. These opportunities include expanding access to infrastructure and data, providing UI faculty development opportunities, strengthening collaborations among faculty, recognizing UI teaching and research excellence, fostering student awareness, developing modular and open-source resources, expanding course-based undergraduate research experiences (CUREs), building curriculum, supporting student professional development and research, and removing financial barriers through funding programs and collaborator support.

Alternative splicing shapes the transcriptome and contributes to each cell's unique identity, but single-cell RNA sequencing (scRNA-seq) has struggled to capture the impact of alternative splicing. We previously showed that low recovery of mRNAs from single cells led to erroneous conclusions about the cell-to-cell variability of alternative splicing. Here, we present a method, Psix, to confidently identify splicing that changes across a landscape of single cells, using a probabilistic model that is robust against the data limitations of scRNA-seq. Its autocorrelation-inspired approach finds patterns of alternative splicing that correspond to patterns of cell identity, such as cell type or developmental stage, without the need for explicit cell clustering, labeling, or trajectory inference. Applying Psix to data that follow the trajectory of mouse brain development, we identify exons whose alternative splicing patterns cluster into modules of coregulation. We show that the exons in these modules are enriched for binding by distinct neuronal splicing factors and that their changes in splicing correspond to changes in expression of these splicing factors. Thus, Psix reveals cell type-dependent splicing patterns and the wiring of the splicing regulatory networks that control them. Our new method will enable scRNA-seq analysis to go beyond transcription to understand the roles of post-transcriptional regulation in determining cell identity.

Intronic polyadenylation (IPA) isoforms, which contain alternative last exons, are widely regulated in various biological processes and by many factors. However, little is known about their cytoplasmic regulation and translational status. In this study, we provide the first evidence that the genome-wide patterns of IPA isoform regulation during a biological process can be very distinct between the transcriptome and translatome, and between the nucleus and cytosol. Indeed, by 3'-seq analyses on breast cancer cells, we show that the genotoxic anticancer drug, doxorubicin, preferentially down-regulates the IPA to the last-exon (IPA:LE) isoform ratio in whole cells (as previously reported) but preferentially up-regulates it in polysomes. We further show that in nuclei, doxorubicin almost exclusively down-regulates the IPA:LE ratio, whereas in the cytosol, it preferentially up-regulates the isoform ratio, as in polysomes. Then, focusing on IPA isoforms that are up-regulated by doxorubicin in the cytosol and highly translated (up-regulated and/or abundant in polysomes), we identify several IPA isoforms that promote cell survival to doxorubicin. Mechanistically, by using an original approach of condition- and compartment-specific CLIP-seq (CCS-iCLIP) to analyze ELAVL1-RNA interactions in the nucleus and cytosol in the presence and absence of doxorubicin, as well as 3'-seq analyses upon ELAVL1 depletion, we show that the RNA-binding protein ELAVL1 mediates both nuclear down-regulation and cytosolic up-regulation of the IPA:LE isoform ratio in distinct sets of genes in response to doxorubicin. Altogether, these findings reveal differential regulation of the IPA:LE isoform ratio across subcellular compartments during drug response and its coordination by an RNA-binding protein.

Massively parallel reporter assays (MPRAs) test the capacity of putative gene regulatory elements to drive transcription on a genome-wide scale. Most gene regulatory activity occurs within accessible chromatin, and recently described methods have combined assays that capture these regions, such as assay for transposase-accessible chromatin using sequencing (ATAC-seq), with self-transcribing active regulatory region sequencing (STARR-seq) to selectively assay the regulatory potential of accessible DNA (ATAC-STARR-seq). Here, we report an integrated approach that quantifies activating and silencing regulatory activity, chromatin accessibility, and transcription factor (TF) occupancy with one assay using ATAC-STARR-seq. Our strategy, including important updates to the ATAC-STARR-seq assay and workflow, enabled high-resolution testing of ~50 million unique DNA fragments tiling ~101,000 accessible chromatin regions in human lymphoblastoid cells. We discovered that 30% of all accessible regions contain an activator, a silencer or both. Although few MPRA studies have explored silencing activity, we demonstrate silencers occur at similar frequencies to activators, and they represent a distinct functional group enriched for unique TF motifs and repressive histone modifications. We further show that Tn5 cut-site frequencies are retained in the ATAC-STARR plasmid library compared to standard ATAC-seq, enabling TF occupancy to be ascertained from ATAC-STARR data. With this approach, we found that activators and silencers cluster by distinct TF footprint combinations and these groups of activity represent different gene regulatory networks of immune cell function. Altogether, these data highlight the multi-layered capabilities of ATAC-STARR-seq to comprehensively investigate the regulatory landscape of the human genome all from a single DNA fragment source.

Transcriptomic diversity greatly contributes to the fundamentals of disease, lineage-specific biology, and environmental adaptation. However, much of the actual isoform repertoire contributing to shaping primate evolution remains unknown. Here, we combined deep long- and short-read sequencing complemented with mass spectrometry proteomics in a panel of lymphoblastoid cell lines (LCLs) from human, three other great apes, and rhesus macaque, producing the largest full-length isoform catalog in primates to date. Around half of the captured isoforms are not annotated in their reference genomes, significantly expanding the gene models in primates. Furthermore, our comparative analyses unveil hundreds of transcriptomic innovations and isoform usage changes related to immune function and immunological disorders. The confluence of these evolutionary innovations with signals of positive selection and their limited impact in the proteome points to changes in alternative splicing in genes involved in immune response as an important target of recent regulatory divergence in primates.