arXiv subjects

Bonnie Berger

Publications and source records attributed to Bonnie Berger.

35 records · Page 2Linked to original sources

arXiv2020

Reconstructing continuous distributions of 3D protein structure from cryo-EM images

Cryo-electron microscopy (cryo-EM) is a powerful technique for determining the structure of proteins and other macromolecular complexes at near-atomic resolution. In single particle cryo-EM, the central problem is to reconstruct the three-dimensional structure of a macromolecule from $10^{4-7}$ noisy and randomly oriented two-dimensional projections. However, the imaged protein complexes may exhibit structural variability, which complicates reconstruction and is typically addressed using discrete clustering approaches that fail to capture the full range of protein dynamics. Here, we introduce a novel method for cryo-EM reconstruction that extends naturally to modeling continuous generative factors of structural heterogeneity. This method encodes structures in Fourier space using coordinate-based deep neural networks, and trains these networks from unlabeled 2D cryo-EM images by combining exact inference over image orientation with variational inference for structural heterogeneity. We demonstrate that the proposed method, termed cryoDRGN, can perform ab initio reconstruction of 3D protein complexes from simulated and real 2D cryo-EM image data. To our knowledge, cryoDRGN is the first neural network-based approach for cryo-EM reconstruction and the first end-to-end method for directly reconstructing continuous ensembles of protein structures from cryo-EM images.

q-bio.QM↗

arXiv2019

Learning protein sequence embeddings using information from structure

Inferring the structural properties of a protein from its amino acid sequence is a challenging yet important problem in biology. Structures are not known for the vast majority of protein sequences, but structure is critical for understanding function. Existing approaches for detecting structural similarity between proteins from sequence are unable to recognize and exploit structural patterns when sequences have diverged too far, limiting our ability to transfer knowledge between structurally related proteins. We newly approach this problem through the lens of representation learning. We introduce a framework that maps any protein sequence to a sequence of vector embeddings --- one per amino acid position --- that encode structural information. We train bidirectional long short-term memory (LSTM) models on protein sequences with a two-part feedback mechanism that incorporates information from (i) global structural similarity between proteins and (ii) pairwise residue contact maps for individual proteins. To enable learning from structural similarity information, we define a novel similarity measure between arbitrary-length sequences of vector embeddings based on a soft symmetric alignment (SSA) between them. Our method is able to learn useful position-specific embeddings despite lacking direct observations of position-level correspondence between sequences. We show empirically that our multi-task framework outperforms other sequence-based methods and even a top-performing structure-based alignment method when predicting structural similarity, our goal. Finally, we demonstrate that our learned embeddings can be transferred to other protein sequence problems, improving the state-of-the-art in transmembrane domain prediction.

cs.LG↗

arXiv2019

Explicitly disentangling image content from translation and rotation with spatial-VAE

Given an image dataset, we are often interested in finding data generative factors that encode semantic content independently from pose variables such as rotation and translation. However, current disentanglement approaches do not impose any specific structure on the learned latent representations. We propose a method for explicitly disentangling image rotation and translation from other unstructured latent factors in a variational autoencoder (VAE) framework. By formulating the generative model as a function of the spatial coordinate, we make the reconstruction error differentiable with respect to latent translation and rotation parameters. This formulation allows us to train a neural network to perform approximate inference on these latent variables while explicitly constraining them to only represent rotation and translation. We demonstrate that this framework, termed spatial-VAE, effectively learns latent representations that disentangle image rotation and translation from content and improves reconstruction over standard VAEs on several benchmark datasets, including applications to modeling continuous 2-D views of proteins from single particle electron microscopy and galaxies in astronomical images.

cs.CV↗

arXiv2018

Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographs

Cryo-electron microscopy (cryoEM) is an increasingly popular method for protein structure determination. However, identifying a sufficient number of particles for analysis (often >100,000) can take months of manual effort. Current computational approaches are limited by high false positive rates and require significant ad-hoc post-processing, especially for unusually shaped particles. To address this shortcoming, we develop Topaz, an efficient and accurate particle picking pipeline using neural networks trained with few labeled particles by newly leveraging the remaining unlabeled particles through the framework of positive-unlabeled (PU) learning. Remarkably, despite using minimal labeled particles, Topaz allows us to improve reconstruction resolution by up to 0.15 Å over published particles on three public cryoEM datasets without any post-processing. Furthermore, we show that our novel generalized-expectation criteria approach to PU learning outperforms existing general PU learning approaches when applied to particle detection, especially for challenging datasets of non-globular proteins. We expect Topaz to be an essential component of cryoEM analysis.

q-bio.QM↗

arXiv2018

Fast Characterization of Segmental Duplications in Genome Assemblies

Segmental duplications (SDs), or low-copy repeats (LCR), are segments of DNA greater than 1 Kbp with high sequence identity that are copied to other regions of the genome. SDs are among the most important sources of evolution, a common cause of genomic structural variation, and several are associated with diseases of genomic origin. Despite their functional importance, SDs present one of the major hurdles for de novo genome assembly due to the ambiguity they cause in building and traversing both state-of-the-art overlap-layout-consensus and de Bruijn graphs. This causes SD regions to be misassembled, collapsed into a unique representation, or completely missing from assembled reference genomes for various organisms. In turn, this missing or incorrect information limits our ability to fully understand the evolution and the architecture of the genomes. Despite the essential need to accurately characterize SDs in assemblies, there is only one tool that has been developed for this purpose, called Whole Genome Assembly Comparison (WGAC). WGAC is comprised of several steps that employ different tools and custom scripts, which makes it difficult and time consuming to use. Thus there is still a need for algorithms to characterize within-assembly SDs quickly, accurately, and in a user friendly manner. Here we introduce a SEgmental Duplication Evaluation Framework (SEDEF) to rapidly detect SDs through sophisticated filtering strategies based on Jaccard similarity and local chaining. We show that SEDEF accurately detects SDs while maintaining substantial speed up over WGAC that translates into practical run times of minutes instead of weeks. Notably, our algorithm captures up to 25% pairwise error between segments, where previous studies focused on only 10%, allowing us to more deeply track the evolutionary history of the genome. SEDEF is available at https://github.com/vpc-ccg/sedef

cs.DS↗

arXiv2018

Large-Margin Classification in Hyperbolic Space

Representing data in hyperbolic space can effectively capture latent hierarchical relationships. With the goal of enabling accurate classification of points in hyperbolic space while respecting their hyperbolic geometry, we introduce hyperbolic SVM, a hyperbolic formulation of support vector machine classifiers, and elucidate through new theoretical work its connection to the Euclidean counterpart. We demonstrate the performance improvement of hyperbolic SVM for multi-class prediction tasks on real-world complex networks as well as simulated datasets. Our work allows analytic pipelines that take the inherent hyperbolic geometry of the data into account in an end-to-end fashion without resorting to ill-fitting tools developed for Euclidean space.

cs.LG↗

arXiv2016

Enabling Privacy-Preserving GWAS in Heterogeneous Human Populations

The projected increase of genotyping in the clinic and the rise of large genomic databases has led to the possibility of using patient medical data to perform genomewide association studies (GWAS) on a larger scale and at a lower cost than ever before. Due to privacy concerns, however, access to this data is limited to a few trusted individuals, greatly reducing its impact on biomedical research. Privacy preserving methods have been suggested as a way of allowing more people access to this precious data while protecting patients. In particular, there has been growing interest in applying the concept of differential privacy to GWAS results. Unfortunately, previous approaches for performing differentially private GWAS are based on rather simple statistics that have some major limitations. In particular, they do not correct for population stratification, a major issue when dealing with the genetically diverse populations present in modern GWAS. To address this concern we introduce a novel computational framework for performing GWAS that tailors ideas from differential privacy to protect private phenotype information, while at the same time correcting for population stratification. This framework allows us to produce privacy preserving GWAS results based on two of the most commonly used GWAS statistics: EIGENSTRAT and linear mixed model (LMM) based statistics. We test our differentially private statistics, PrivSTRAT and PrivLMM, on both simulated and real GWAS datasets and find that they are able to protect privacy while returning meaningful GWAS results.

q-bio.QM↗

arXiv2016

Low-density locality-sensitive hashing boosts metagenomic binning

Metagenomic binning is an essential task in analyzing metagenomic sequence datasets. To analyze structure or function of microbial communities from environmental samples, metagenomic sequence fragments are assigned to their taxonomic origins. Although sequence alignment algorithms can readily be used and usually provide high-resolution alignments and accurate binning results, the computational cost of such alignment-based methods becomes prohibitive as metagenomic datasets continue to grow. Alternative compositional-based methods, which exploit sequence composition by profiling local short k-mers in fragments, are often faster but less accurate than alignment-based methods. Inspired by the success of linear error correcting codes in noisy channel communication, we introduce Opal, a fast and accurate novel compositional-based binning method. It incorporates ideas from Gallager's low-density parity-check code to design a family of compact and discriminative locality-sensitive hashing functions that encode long-range compositional dependencies in long fragments. By incorporating the Gallager LSH functions as features in a simple linear SVM, Opal provides fast, accurate and robust binning for datasets consisting of a large number of species, even with mutations and sequencing errors. Opal not only performs up to two orders of magnitude faster than BWA, an alignment-based binning method, but also achieves improved binning accuracy and robustness to sequencing errors. Opal also outperforms models built on traditional k-mer profiles in terms of robustness and accuracy. Finally, we demonstrate that we can effectively use Opal in the "coarse search" stage of a compressive genomics pipeline to identify a much smaller candidate set of taxonomic origins for a subsequent alignment-based method to analyze, thus providing metagenomic binning with high scalability, high accuracy and high resolution.

q-bio.QM↗

arXiv2015

Entropy-scaling search of massive biological data

Many datasets exhibit a well-defined structure that can be exploited to design faster search tools, but it is not always clear when such acceleration is possible. Here, we introduce a framework for similarity search based on characterizing a dataset's entropy and fractal dimension. We prove that searching scales in time with metric entropy (number of covering hyperspheres), if the fractal dimension of the dataset is low, and scales in space with the sum of metric entropy and information-theoretic entropy (randomness of the data). Using these ideas, we present accelerated versions of standard tools, with no loss in specificity and little loss in sensitivity, for use in three domains---high-throughput drug screening (Ammolite, 150x speedup), metagenomics (MICA, 3.5x speedup of DIAMOND [3,700x BLASTX]), and protein structure search (esFragBag, 10x speedup of FragBag). Our framework can be used to achieve "compressive omics," and the general theory can be readily applied to data science problems outside of biology.

cs.DS↗

arXiv2015

iTreePack: Protein Complex Side-Chain Packing by Dual Decomposition

Protein side-chain packing is a critical component in obtaining the 3D coordinates of a structure and drug discovery. Single-domain protein side-chain packing has been thoroughly studied. A major challenge in generalizing these methods to protein complexes is that they, unlike monomers, often have very large treewidth, and thus algorithms such as TreePack cannot be directly applied. To address this issue, SCWRL4 treats the complex effectively as a monomer, heuristically excluding weak interactions to decrease treewidth; as a result, SCWRL4 generates poor packings on protein interfaces. To date, few side-chain packing methods exist that are specifically designed for protein complexes. In this paper, we introduce a method, iTreePack, which solves the side-chain packing problem for complexes by using a novel combination of dual decomposition and tree decomposition. In particular, iTreePack overcomes the problem of large treewidth by decomposing a protein complex into smaller subgraphs and novelly reformulating the complex side-chain packing problem as a dual relaxation problem; this allows us to solve the side-chain packing of each small subgraph separately using tree-decomposition. A projected subgradient algorithm is applied to enforcing the consistency among the side-chain packings of all the small subgraphs. Computational results demonstrate that our iTreePack program outperforms SCWRL4 on protein complexes. In particular, iTreePack places side-chain atoms much more accurately on very large complexes, which constitute a significant portion of protein-protein interactions. Moreover, the advantage of iTreePack over SCWRL4 increases with respect to the treewidth of a complex. Even for monomeric proteins, iTreePack is much more efficient than SCWRL and slightly more accurate.

q-bio.BM↗

arXiv2015

Diffusion Component Analysis: Unraveling Functional Topology in Biological Networks

Complex biological systems have been successfully modeled by biochemical and genetic interaction networks, typically gathered from high-throughput (HTP) data. These networks can be used to infer functional relationships between genes or proteins. Using the intuition that the topological role of a gene in a network relates to its biological function, local or diffusion based "guilt-by-association" and graph-theoretic methods have had success in inferring gene functions. Here we seek to improve function prediction by integrating diffusion-based methods with a novel dimensionality reduction technique to overcome the incomplete and noisy nature of network data. In this paper, we introduce diffusion component analysis (DCA), a framework that plugs in a diffusion model and learns a low-dimensional vector representation of each node to encode the topological properties of a network. As a proof of concept, we demonstrate DCA's substantial improvement over state-of-the-art diffusion-based approaches in predicting protein function from molecular interaction networks. Moreover, our DCA framework can integrate multiple networks from heterogeneous sources, consisting of genomic information, biochemical experiments and other resources, to even further improve function prediction. Yet another layer of performance gain is achieved by integrating the DCA framework with support vector machines that take our node vector representations as features. Overall, our DCA framework provides a novel representation of nodes in a network that can be used as a plug-in architecture to other machine learning algorithms to decipher topological properties of and obtain novel insights into interactomes.

q-bio.MN↗

arXiv2014

Ancient human genomes suggest three ancestral populations for present-day Europeans

We sequenced genomes from a $\sim$7,000 year old early farmer from Stuttgart in Germany, an $\sim$8,000 year old hunter-gatherer from Luxembourg, and seven $\sim$8,000 year old hunter-gatherers from southern Sweden. We analyzed these data together with other ancient genomes and 2,345 contemporary humans to show that the great majority of present-day Europeans derive from at least three highly differentiated populations: West European Hunter-Gatherers (WHG), who contributed ancestry to all Europeans but not to Near Easterners; Ancient North Eurasians (ANE), who were most closely related to Upper Paleolithic Siberians and contributed to both Europeans and Near Easterners; and Early European Farmers (EEF), who were mainly of Near Eastern origin but also harbored WHG-related ancestry. We model these populations' deep relationships and show that EEF had $\sim$44% ancestry from a "Basal Eurasian" lineage that split prior to the diversification of all other non-African lineages.

q-bio.PE↗

arXiv2013

Ancient west Eurasian ancestry in southern and eastern Africa

The history of southern Africa involved interactions between indigenous hunter-gatherers and a range of populations that moved into the region. Here we use genome-wide genetic data to show that there are at least two admixture events in the history of Khoisan populations (southern African hunter-gatherers and pastoralists who speak non-Bantu languages with click consonants). One involved populations related to Niger-Congo-speaking African populations, and the other introduced ancestry most closely related to west Eurasian (European or Middle Eastern) populations. We date this latter admixture event to approximately 900-1,800 years ago, and show that it had the largest demographic impact in Khoisan populations that speak Khoe-Kwadi languages. A similar signal of west Eurasian ancestry is present throughout eastern Africa. In particular, we also find evidence for two admixture events in the history of Kenyan, Tanzanian, and Ethiopian populations, the earlier of which involved populations related to west Eurasians and which we date to approximately 2,700 - 3,300 years ago. We reconstruct the allele frequencies of the putative west Eurasian population in eastern Africa, and show that this population is a good proxy for the west Eurasian ancestry in southern Africa. The most parsimonious explanation for these findings is that west Eurasian ancestry entered southern Africa indirectly through eastern Africa.

q-bio.PE↗

arXiv2013

Efficient moment-based inference of admixture parameters and sources of gene flow

The recent explosion in available genetic data has led to significant advances in understanding the demographic histories of and relationships among human populations. It is still a challenge, however, to infer reliable parameter values for complicated models involving many populations. Here we present MixMapper, an efficient, interactive method for constructing phylogenetic trees including admixture events using single nucleotide polymorphism (SNP) genotype data. MixMapper implements a novel two-phase approach to admixture inference using moment statistics, first building an unadmixed scaffold tree and then adding admixed populations by solving systems of equations that express allele frequency divergences in terms of mixture parameters. Importantly, all features of the model, including topology, sources of gene flow, branch lengths, and mixture proportions, are optimized automatically from the data and include estimates of statistical uncertainty. MixMapper also uses a new method to express branch lengths in easily interpretable drift units. We apply MixMapper to recently published data for HGDP individuals genotyped on a SNP array designed especially for use in population genetics studies, obtaining confident results for 30 populations, 20 of them admixed. Notably, we confirm a signal of ancient admixture in European populations---including previously undetected admixture in Sardinians and Basques---involving a proportion of 20--40% ancient northern Eurasian ancestry.

q-bio.PE↗

arXiv2013

Inferring Admixture Histories of Human Populations Using Linkage Disequilibrium

Long-range migrations and the resulting admixtures between populations have been important forces shaping human genetic diversity. Most existing methods for detecting and reconstructing historical admixture events are based on allele frequency divergences or patterns of ancestry segments in chromosomes of admixed individuals. An emerging new approach harnesses the exponential decay of admixture-induced linkage disequilibrium (LD) as a function of genetic distance. Here, we comprehensively develop LD-based inference into a versatile tool for investigating admixture. We present a new weighted LD statistic that can be used to infer mixture proportions as well as dates with fewer constraints on reference populations than previous methods. We define an LD-based three-population test for admixture and identify scenarios in which it can detect admixture events that previous formal tests cannot. We further show that we can uncover phylogenetic relationships among populations by comparing weighted LD curves obtained using a suite of references. Finally, we describe several improvements to the computation and fitting of weighted LD curves that greatly increase the robustness and speed of the calculations. We implement all of these advances in a software package, ALDER, which we validate in simulations and apply to test for admixture among all populations from the Human Genome Diversity Project (HGDP), highlighting insights into the admixture history of Central African Pygmies, Sardinians, and Japanese.

q-bio.PE↗

arXiv2012

Reconstructing Roma history from genome-wide data

The Roma people, living throughout Europe, are a diverse population linked by the Romani language and culture. Previous linguistic and genetic studies have suggested that the Roma migrated into Europe from South Asia about 1000-1500 years ago. Genetic inferences about Roma history have mostly focused on the Y chromosome and mitochondrial DNA. To explore what additional information can be learned from genome-wide data, we analyzed data from six Roma groups that we genotyped at hundreds of thousands of single nucleotide polymorphisms (SNPs). We estimate that the Roma harbor about 80% West Eurasian ancestry-deriving from a combination of European and South Asian sources- and that the date of admixture of South Asian and European ancestry was about 850 years ago. We provide evidence for Eastern Europe being a major source of European ancestry, and North-west India being a major source of the South Asian ancestry in the Roma. By computing allele sharing as a measure of linkage disequilibrium, we estimate that the migration of Roma out of the Indian subcontinent was accompanied by a severe founder event, which we hypothesize was followed by a major demographic expansion once the population arrived in Europe.

q-bio.PE↗

arXiv2012

The genetic prehistory of southern Africa

Southern and eastern African populations that speak non-Bantu languages with click consonants are known to harbour some of the most ancient genetic lineages in humans, but their relationships are poorly understood. Here, we report data from 23 populations analyzed at over half a million single nucleotide polymorphisms, using a genome-wide array designed for studying human history. The southern African Khoisan fall into two genetic groups, loosely corresponding to the northwestern and southeastern Kalahari, which we show separated within the last 30,000 years. We find that all individuals derive at least a few percent of their genomes from admixture with non-Khoisan populations that began approximately 1,200 years ago. In addition, the east African Hadza and Sandawe derive a fraction of their ancestry from admixture with a population related to the Khoisan, supporting the hypothesis of an ancient link between southern and eastern Africa

q-bio.PE↗