Search

The current human reference genome, GRCh38, represents over 20 years of effort to generate a high-quality assembly, which has benefitted society1,2. However, it still has many gaps and errors, and does not represent a biological genome as it is a blend of multiple individuals3,4. Recently, a high-quality telomere-to-telomere reference, CHM13, was generated with the latest long-read technologies, but it was derived from a hydatidiform mole cell line with a nearly homozygous genome5. To address these limitations, the Human Pangenome Reference Consortium formed with the goal of creating high-quality, cost-effective, diploid genome assemblies for a pangenome reference that represents human genetic diversity6. Here, in our first scientific report, we determined which combination of current genome sequencing and assembly approaches yield the most complete and accurate diploid genome assembly with minimal manual curation. Approaches that used highly accurate long reads and parent–child data with graph-based haplotype phasing during assembly outperformed those that did not. Developing a combination of the top-performing methods, we generated our first high-quality diploid reference assembly, containing only approximately four gaps per chromosome on average, with most chromosomes within ±1% of the length of CHM13. Nearly 48% of protein-coding genes have non-synonymous amino acid changes between haplotypes, and centromeric regions showed the highest diversity. Our findings serve as a foundation for assembling near-complete diploid human genomes at scale for a pangenome reference to capture global genetic variation from single nucleotides to structural rearrangements.

Sequence divergence ofcis-regulatory elements drives species-specific traits, but how this manifests in the evolution of the neocortex at the molecular and cellular level remains to be elucidated. We investigated the gene regulatory programs in the primary motor cortex of human, macaque, marmoset, and mouse with single-cell multiomics assays, generating gene expression, chromatin accessibility, DNA methylome, and chromosomal conformation profiles from a total of over 180,000 cells. For each modality, we determined species-specific, divergent, and conserved gene expression and epigenetic features at multiple levels. We find that cell type-specific gene expression evolves more rapidly than broadly expressed genes and that epigenetic status at distal candidatecis-regulatory elements (cCREs) evolves faster than promoters. Strikingly, transposable elements (TEs) contribute to nearly 80% of the human-specific cCREs in cortical cells. Through machine learning, we develop sequence-based predictors of cCREs in different species and demonstrate that the genomic regulatory syntax is highly preserved from rodents to primates. Lastly, we show that epigenetic conservation combined with sequence similarity helps uncover functionalcis-regulatory elements and enhances our ability to interpret genetic variants contributing to neurological disease and traits.

The current human reference genome, GRCh38, represents over 20 years of effort to generate a high-quality assembly, which has benefitted society1,2. However, it still has many gaps and errors, and does not represent a biological genome as it is a blend of multiple individuals3,4. Recently, a high-quality telomere-to-telomere reference, CHM13, was generated with the latest long-read technologies, but it was derived from a hydatidiform mole cell line with a nearly homozygous genome5. To address these limitations, the Human Pangenome Reference Consortium formed with the goal of creating high-quality, cost-effective, diploid genome assemblies for a pangenome reference that represents human genetic diversity6. Here, in our first scientific report, we determined which combination of current genome sequencing and assembly approaches yield the most complete and accurate diploid genome assembly with minimal manual curation. Approaches that used highly accurate long reads and parent-child data with graph-based haplotype phasing during assembly outperformed those that did not. Developing a combination of the top-performing methods, we generated our first high-quality diploid reference assembly, containing only approximately four gaps per chromosome on average, with most chromosomes within ±1% of the length of CHM13. Nearly 48% of protein-coding genes have non-synonymous amino acid changes between haplotypes, and centromeric regions showed the highest diversity. Our findings serve as a foundation for assembling near-complete diploid human genomes at scale for a pangenome reference to capture global genetic variation from single nucleotides to structural rearrangements.

The human reference genome is the most widely used resource in human genetics and is due for a major update. Its current structure is a linear composite of merged haplotypes from more than 20 people, with a single individual comprising most of the sequence. It contains biases and errors within a framework that does not represent global human genomic variation. A high-quality reference with global representation of common variants, including single-nucleotide variants, structural variants and functional elements, is needed. The Human Pangenome Reference Consortium aims to create a more sophisticated and complete human reference genome with a graph-based, telomere-to-telomere representation of global genomic diversity. Here we leverage innovations in technology, study design and global partnerships with the goalof constructing the highest-possible quality human pangenome reference. Our goal is toimprove data representation and streamline analyses to enable routine assembly of complete diploid genomes. With attention to ethical frameworks, the human pangenome reference will contain a more accurate and diverse representation of global genomic variation, improve gene-disease association studies across populations, expand the scope of genomics research to the most repetitive and polymorphic regions of the genome, and serve as the ultimate genetic resource for future biomedical research and precision medicine.

The human reference genome is the most widely-used resource in human genetics and is due for a major update. Its current structure is a linear composite of merged haplotypes from more than 20 people, with a single individual comprising most of the sequence. It contains biases and errors within a framework that does not represent global human genomic variation. A high-quality reference with global representation of common variants, including single nucleotide variants (SNVs), structural variants (SVs), and functional elements is needed. The Human Pangenome Reference Consortium (HPRC) aims to create a more sophisticated and complete human reference genome with a graph-based, telomere-to-telomere representation of global genomic diversity. We will leverage innovations in technology, study design, and global partnerships to construct the highest-possible quality human pangenome reference. Our effort will improve data representation and streamline analyses to enable routine assembly of complete diploid genomes. With attention to ethical frameworks, the human pangenome reference will contain a more accurate and diverse representation of global genomic variation, improve gene-disease association studies across populations, expand the scope of genomics research to the most repetitive and polymorphic regions of the genome, and serve as the ultimate genetic resource for future biomedical research and precision medicine.

Federated learning (FL) is a burgeoning distributed machine learning framework where a central parameter server (PS) coordinates many local users to train a globally consistent model. Conventional federated learning inevitably relies on a centralized topology with a PS. As a result, it will paralyze once the PS fails. To alleviate such a single point failure, especially on the PS, some existing work has provided decentralized FL (DFL) implementations like CDSGD and D-PSGD to facilitate FL in a decentralized topology. However, there are still some problems with these methods, e.g., significant divergence between users' final models in CDSGD and a network-wide model average necessity in D-PSGD. In order to solve these deficiency, this paper devises a new DFL implementation coined as DACFL, where each user trains its model using its own training data and exchanges the intermediate models with its neighbors through a symmetric and doubly stochastic matrix. The DACFL treats the progress of each user's local training as a discrete-time process and employs a first order dynamic average consensus (FODAC) method to track the \textit{average model} in the absence of the PS. In this paper, we also provide a theoretical convergence analysis of DACFL on the premise of i.i.d data to strengthen its rationality. The experimental results on MNIST, Fashion-MNIST and CIFAR-10 validate the feasibility of our solution in both time-invariant and time-varying network topologies, and declare that DACFL outperforms D-PSGD and CDSGD in most cases.

Additional file 3:. Review history

Additional file 1: Supplemental Table Legends and Figures S1-S11

Uterine carcinosarcoma (UCS) is a form of endometrial cancer simultaneously exhibiting carcinomatous and sarcomatous elements, but the underlying molecular and epigenetic basis of this disease is poorly understood. We generated complete DNA methylomes for both the carcinomatous and the sarcomatous components of three UCS samples separated by laser capture microdissection and compared DNA methylomes of UCS with those of normal endometrium as well as methylomes derived from endometrioid carcinoma, serous endometrial carcinoma, and endometrial stromal sarcoma. We identified epigenetic lesions specific to carcinosarcoma and specific to its two components. Hallmarks of DNA methylation abnormalities in UCS included global hypomethylation, especially in repetitive elements, and hypermethylation of tumor suppressor gene promoters. Among these, aberrant DNA methylation of MIR200 genes is a key feature of UCS. The carcinoma component of UCS was characterized by hypermethylation of promoters of EMILIN1, NEFM, and CLEC14A, genes that are associated with tumor vascularization. In contrast, DNA methylation changes of PKP3, FAM83F, and TCP11 were more characteristic of the sarcoma components. Our findings highlight the epigenetic signatures that distinguish the two components of UCS, providing a valuable resource for investigation of this disease.

Recent advancements in sequencing-based DNA methylation profiling methods provide an unprecedented opportunity to map complete DNA methylomes. These include whole-genome bisulfite sequencing (WGBS, MethylC-seq, or BS-seq), reduced-representation bisulfite sequencing (RRBS), and enrichment-based methods such as MeDIP-seq, MBD-seq, and MRE-seq. These methods yield largely comparable results but differ significantly in extent of genomic CpG coverage, resolution, quantitative accuracy, and cost, at least while using current algorithms to interrogate the data. None of these existing methods provides single-CpG resolution, comprehensive genome-wide coverage, and cost feasibility for a typical laboratory. We introduce methylCRF, a novel conditional random fields-based algorithm that integrates methylated DNA immunoprecipitation (MeDIP-seq) and methylation-sensitive restriction enzyme (MRE-seq) sequencing data to predict DNA methylation levels at single-CpG resolution. Our method is a combined computational and experimental strategy to produce DNA methylomes of all 28 million CpGs in the human genome for a fraction (

In order to better reflect the response of natural geographical environment factors, such as hydrology, climate and soil, to the changes of land use/land cover, land cover classification using multi-temporal and multi-spectral remote sensing observation data has been an effective technique. This paper adopts a multi-dimension classification method to classify the vegetation cover of the Yellow River Basin (YRB) in China by using NOAA/AVHRR remote sensing image as data source. Better classification resolution, which Mill provide scientific basis to study on increasingly prominent ecological-environmental problems, hydrology and water resources of the YRB, is acquired by this complex classification method. With the support of geography information system (GIS) and remote sensing (RS), this paper makes use of principal component transformation (KL transformation) to classify the vegetation cover using the Normalized Difference Vegetation Index (NDVI) data of NOAA/AVHRR remote sensing image as data source with spatial and temporal resolution at 1 km times 1 km and a vegetation growth cycle. Principal component transformation is first carried out on 3 normalized images of bio-temperature (BT), potential evapotranspiration rate (PER) and annual precipitation (P) which are the controlling factors on vegetation distribution pattern and the first principal component is chosen as the first classification vector of the digital image of the vegetation cover classification of the YRB. Because the basin topography is complex and underlaying surface condition has great impact on vegetation growth, Digital Elevation Model (DEM) image of the YRB became the second classification vector. Then, the first 3 principal components from the time series of the 12 monthly NDVI images are combined with the first and the second classification vectors to consist of the integrated image with 5 bands. After integrating the multidimension information, the vegetation cover of the YRB is finally classified into two grads with iterative self-organizing data analysis technique A (ISODATA) and unsupervised classification method, including 8 vegetation types and 25 vegetation subtypes