Biology subjects

Berger, B.

Publications and source records attributed to Berger, B..

At least 37 records · Page 2Linked to original sources

bioRxiv2023

Strain Tracking with Uncertainty Quantification

The ability to detect and quantify microbiota over time has a plethora of clinical, basic science, and public health applications. One of the primary means of tracking microbiota is through sequencing technologies. When the microorganism of interest is well characterized or known a priori, targeted sequencing is often used. In many applications, however, untargeted bulk (shotgun) sequencing is more appropriate; for instance, the tracking of infection transmission events and nucleotide variants across multiple genomic loci, or studying the role of multiple genes in a particular phenotype. Given these applications, and the observation that pathogens (e.g. Clostridioides difficile, Escherichia coli, Salmonella enterica) and other taxa of interest can reside at low relative abundance in the gastrointestinal tract, there is a critical need for algorithms that accurately track low-abundance taxa with strain level resolution. Here we present a sequence quality- and time-aware model, ChronoStrain, that introduces uncertainty quantification to gauge low-abundance species and significantly outperforms the current state-of-the-art on both real and synthetic data. ChronoStrain leverages sequences quality scores and the samples temporal information to produce a probability distribution over abundance trajectories for each strain tracked in the model. We demonstrate Chronostrains improved performance in capturing post-antibiotic Escherichia coli strain blooms among women with recurrent urinary tract infections (UTIs) from the UTI Microbiome (UMB) Project. Other strain tracking models on the same data either show inconsistent temporal colonization or can only track consistently using very coarse groupings. In contrast, our probabilistic outputs can reveal the relationship between low-confidence strains present in the sample that cannot be reliably assigned a single reference label (either due to poor coverage or novelty) while simultaneously calling high-confidence strains that can be unambiguously assigned a label. We also analyze samples from the Early Life Microbiota Colonisation (ELMC) Study demonstrating the algorithms ability to correctly identify Enterococcus faecalis strains using paired sample isolates as validation.

bioinformatics↗

bioRxiv2022

mapquik: Efficient low-divergence mapping of long reads in minimizer space

DNA sequencing data continues to progress towards longer reads with increasingly lower sequencing error rates. We focus on the critical problem of mapping, or aligning, low-divergence sequences from long reads (PacBio HiFi) to a reference genome, which poses challenges in terms of accuracy and computational resources when using cutting-edge read mapping approaches that are designed for all types of alignments. A natural idea would be to optimize efficiency with longer seeds to reduce the probability of extraneous matches; however, contiguous exact seeds quickly reach a sensitivity limit. We introduce mapquik, a novel strategy that creates accurate longer seeds by anchoring alignments through matches of k consecutively-sampled minimizers (k-min-mers) and only indexing k-min-mers that occur once in the reference genome, thereby unlocking ultra-fast mapping while retaining high sensitivity. We demonstrate that mapquik significantly accelerates the seeding and chaining steps -- fundamental bottlenecks to read mapping -- for both the human and maize genomes with > 96% sensitivity and near-perfect specificity. On the human genome, mapquik achieves a 30x speed-up over the state-of-the-art tool minimap2, and on the maize genome, a 350x speed-up over minimap2, making mapquik the fastest mapper to date. These accelerations are enabled not only by minimizer-space seeding but also a novel heuristic [O](n) pseudo-chaining algorithm, which improves over the long-standing [O](n log n) bound. Minimizer-space computation builds the foundation for achieving real-time analysis of long-read sequencing data.

bioinformatics↗

bioRxiv2022

Optimal parameters for Rapid invisible Frequency Tagging using MEG

Frequency tagging has been demonstrated to be a useful tool for identifying representational-specific neuronal activity in the auditory and visual domains. However, the slow flicker (<30Hz) applied in conventional frequency tagging studies is highly visible and might entrain endogenous neuronal oscillations. Hence, stimulation at faster frequencies that is much less visible and does not interfere with endogenous brain oscillatory activity is a promising new tool. In this study, we set out to examine the optimal stimulation parameters of rapid invisible frequency tagging (RFT/RIFT) with magnetoencephalography (MEG) by quantifying the effects of stimulation frequency, size and position of the flickering patch. Rapid frequency tagging (RFT) using flickers above 50 Hz results in almost invisible stimulation which does not interfere with slower endogenous oscillations; however, the signal is weaker as compared to tagging at slower frequencies so the optimal parameters of stimulation delivery are crucial. The here presented results examining the frequency range between 60Hz and 96Hz suggest that RFT induces brain responses with decreasing strength up to about 84Hz. In addition, even at the smallest flicker patch (2{degrees}) focally presented RFT induces a significant oscillatory brain signal at the stimulation frequency (66Hz); however, the elicited response increases with patch size. While focal RFT presentation elicits the strongest response, off-centre presentations do generally mainly elicit a measureable response if presented below the horizontal midline. The results also revealed considerable individual differences in the neuronal responses of to RFT stimulation. Finally, we discuss the comparison of oscillatory measures (coherence and power) and sensor types (planar gradiometers and magnetometers) in order to achieve optimal outcomes. Based on our extensive findings we set forward concrete recommendations for using rapid frequency tagging in human cognitive neuroscience investigations.

neuroscience↗

bioRxiv2022

Learning the Drug-Target Interaction Lexicon

Sequence-based prediction of drug-target interactions has the potential to accelerate drug discovery by complementing experimental screens. Such computational prediction needs to be generalizable and scalable while remaining sensitive to subtle variations in the inputs. However, current computational techniques fail to simultaneously meet these goals, often sacrificing performance on one to achieve the others. We develop a deep learning model, ConPLex, successfully leveraging the advances in pre-trained protein language models ("PLex") and employing a novel protein-anchored contrastive co-embedding ("Con") to outperform state-of-the-art approaches. ConPLex achieves high accuracy, broad adaptivity to unseen data, and specificity against decoy compounds. It makes predictions of binding based on the distance between learned representations, enabling predictions at the scale of massive compound libraries and the human proteome. Furthermore, ConPLex is interpretable, which enables us to visualize the drug-target lexicon and use embeddings to characterize the function of human cell-surface proteins. We anticipate ConPLex will facilitate novel drug discovery by making highly sensitive and interpretable in-silico drug screening feasible at genome scale. Con-PLex is available open-source at https://github.com/samsledje/ConPLex. Significance StatementIn time and money, one of the most expensive steps of the drug discovery pipeline is the experimental screening of small molecules to see which will bind to a protein target of interest. Therefore, accurate high-throughput computational prediction of drug-target interactions would unlock significant value, guiding and prioritizing promising candidates for experimental screening. We introduce ConPLex, a machine learning method for predicting drug-target binding which achieves state-of-the-art accuracy on many types of targets by using a pre-trained protein language model. The approach co-locates the proteins and the potential drug molecules in a shared feature space while learning to contrast true drugs from similar non-binding "decoy" molecules. ConPLex is extremely fast, which allows it to rapidly shortlist candidates for deeper investigation.

bioinformatics↗

bioRxiv2022

Secure and Federated Genome-Wide Association Studies for Biobank-Scale Datasets

Sharing data across institutions for genome-wide association studies (GWAS) would enhance the discovery of genetic variants linked to health and disease1, 2. However, existing data sharing regulations limit the scope of such collaborations3. Although cryptographic tools for secure computation promise to enable collaborative analysis with formal privacy guarantees, existing approaches either are computationally impractical or support only simplified analyses4-7. We introduce secure federated genome-wide association studies (SF-GWAS), a novel combination of secure computation frameworks and distributed algorithms that empowers efficient and accurate GWAS on private data held by multiple entities while ensuring data confidentiality. SF-GWAS supports the most widely-used GWAS pipelines based on principal component analysis (PCA) or linear mixed models (LMMs). We demonstrate the accuracy and practical runtimes of SF-GWAS on five datasets, including a large UK Biobank cohort of 410K individuals, showcasing an order-of-magnitude improvement in runtime compared to previous work. Our work realizes the power of secure collaborative genomic studies at unprecedented scale.

bioinformatics↗

bioRxiv2022

Adapting protein language models for rapid DTI prediction

We consider the problem of sequence-based drug-target interaction (DTI) prediction, showing that a straightforward deep learning architecture that leverages pre-trained protein language models (PLMs) for protein embedding outperforms state of the art approaches, achieving higher accuracy, expanded generalizability, and an order of magnitude faster training. PLM embeddings are found to contain general information that is especially useful in few-shot (small training data set) and zero-shot instances (unseen proteins or drugs). Additionally, the PLM embeddings can be augmented with features tuned by task-specific pre-training, and we find that these task-specific features are more informative than baseline PLM features. We anticipate such transfer learning approaches will facilitate rapid prototyping of DTI models, especially in low-N scenarios.

bioinformatics↗

bioRxiv2022

Contrasting drugs from decoys

Protein language models (PLMs) have recently been proposed to advance drugtarget interaction (DTI) prediction, and have shown state-of-the-art performance on several standard benchmarks. However, a remaining challenge for all DTI prediction models (including PLM-based ones) is distinguishing true drugs from highly-similar decoys. Leveraging techniques from self-supervised contrastive learning, we introduce a second-generation PLM-based DTI model trained on triplets of proteins, drugs, and decoys (small drug-like molecules that do not bind to the protein). We show that our approach, CON-Plex, improves specificity while maintaining high prediction accuracy and generalizability to new drug classes. CON-Plex maps proteins and drugs to a shared latent space which can be interpreted to identify mutually-compatible classes of proteins and drugs. Data and code are available at https://zenodo.org/record/7127229.

bioinformatics↗

bioRxiv2022

Efficient minimizer orders for large values of k using minimum decycling sets

Minimizers are ubiquitously used in data structures and algorithms for efficient searching, mapping, and indexing of high-throughput DNA sequencing data. Minimizer schemes select a minimum k-mer in every L-long sub-sequence of the target sequence, where minimality is with respect to a predefined k-mer order. Commonly used minimizer orders select more k-mers than necessary and therefore provide limited improvement in runtime and memory usage of downstream analysis tasks. The recently introduced universal k-mer hitting sets produce minimizer orders with fewer selected k-mers. Unfortunately, generating compact universal k-mer hitting sets is currently infeasible for k > 13, and thus cannot help in the many applications that require minimizer orders for larger k. Here, we close the gap of efficient minimizer orders for large values of k by introducing decycling-set-based minimizer orders, new orders based on minimum decycling sets. We show that in practice these new minimizer orders select a number of k-mers comparable to that of minimizer orders based on universal k-mer hitting sets, and can also scale up to larger k. Furthermore, we developed a method that computes the minimizers in a sequence on the fly without keeping the k-mers of a decycling set in memory. This enables the use of these minimizer orders for any value of k. We expect the new orders to improve the runtime and memory usage of algorithms and data structures in high-throughput DNA sequencing analysis.

bioinformatics↗

bioRxiv2022

Unraveling causal gene regulation from the RNA velocity graph using Velorama

Single-cell expression dynamics from differentiation trajectories or RNA velocity have the potential to reveal causal links between transcription factors (TFs) and their target genes in gene regulatory networks (GRNs). However, existing methods either neglect these expression dynamics or require cells to be ordered along a linear pseudotemporal axis, which is incompatible with branching trajectories. We introduce Velorama, an approach to causal GRN inference that represents single-cell differentiation dynamics as a directed acyclic graph (DAG) of cells constructed from pseudotime or RNA velocity measurements. In contrast to previous approaches, Velorama is able to work directly with RNA velocity-based cell-to-cell transition probabilities and enables estimates of TF interaction speeds with their target genes. On a set of synthetic datasets, Velorama substantially outperforms existing approaches, improving area under the precision-recall curve (AUPRC) by 3.7-4.8x over the next best method. Applying Velorama to four RNA velocity datasets, we uncover evidence that the speed of a TFs interactions is tied to its regulatory function. For human corticogenesis, we find slow TFs to be linked to gliomas and co-regulate preferentially with fast TFs, while fast TFs are associated with neuropsychiatric diseases. We expect Velorama to be a critical part of the RNA velocity toolkit for investigating the causal drivers of differentiation and disease. Software availabilityhttps://cb.csail.mit.edu/cb/velorama

bioinformatics↗

bioRxiv2022

Mechanisms of mRNA processing defects in inherited THOC6 intellectual disability syndrome

THOC6 is the genetic basis of autosomal recessive THOC6 Intellectual Disability Syndrome (TIDS). THOC6 facilitates the formation of the Transcription Export complex (TREX) tetramer, composed of four THO monomers. The TREX tetramer supports mammalian mRNA processing that is distinct from yeast TREX dimer functions. Human and mouse TIDS model systems allow novel THOC6-dependent TREX tetramer functions to be investigated. Biallelic loss-of-function (LOF) THOC6 variants do not influence the expression and localization of TREX members in human cells, but our data suggests reduced binding affinity of ALYREF. Impairment of TREX nuclear export functions were not detected in cells with biallelic THOC6 LOF. Instead, mRNA mis-splicing was observed in human and mouse neural tissue, revealing novel insights into THOC6-mediated TREX coordination of mRNA processing. We demonstrate that THOC6 is required for regulation of key signaling pathways in human corticogenesis that dictate the transition from proliferative to neurogenic divisions that may inform TIDS neuropathology.

molecular biology↗

bioRxiv2022

Uncovering structural ensembles from single particle cryo-EM data using cryoDRGN

CryoDRGN is a machine learning system for heterogeneous cryo-EM reconstruction of proteins and protein complexes from single particle cryo-EM data. Central to this approach is a deep generative model for heterogeneous cryo-EM density maps, which we empirically find effectively models both discrete and continuous forms of structural variability. Once trained, cryoDRGN is capable of generating an arbitrary number of 3D density maps, and thus interpreting the resulting ensemble is a challenge. Here, we showcase interactive and automated processing approaches for analyzing cryoDRGN results. Specifically, we detail a step-by-step protocol for analysis of the assembling 50S ribosome dataset (Davis et al., EMPIAR-10076), including preparation of inputs, network training, and visualization of the resulting ensemble of density maps. Additionally, we describe and implement methods to comprehensively analyze and interpret the distribution of volumes with the assistance of an associated atomic model. This protocol is appropriate for structural biologists familiar with processing single particle cryo-EM datasets and with moderate experience programming in Python and navigating Jupyter notebooks. It requires 3-4 days to complete.

biochemistry↗

bioRxiv2022

High-resolution de novo structure prediction from primary sequence

Recent breakthroughs have used deep learning to exploit evolutionary information in multiple sequence alignments (MSAs) to accurately predict protein structures. However, MSAs of homologous proteins are not always available, such as with orphan proteins or fast-evolving proteins like antibodies, and a protein typically folds in a natural setting from its primary amino acid sequence into its three-dimensional structure, suggesting that evolutionary information and MSAs should not be necessary to predict a proteins folded form. Here, we introduce OmegaFold, the first computational method to successfully predict high-resolution protein structure from a single primary sequence alone. Using a new combination of a protein language model that allows us to make predictions from single sequences and a geometry-inspired transformer model trained on protein structures, OmegaFold outperforms RoseTTAFold and achieves similar prediction accuracy to AlphaFold2 on recently released structures. OmegaFold enables accurate predictions on orphan proteins that do not belong to any functionally characterized protein family and antibodies that tend to have noisy MSAs due to fast evolution. Our study fills a much-encountered gap in structure prediction and brings us a step closer to understanding protein folding in nature.

bioinformatics↗

bioRxiv2022

Optimal transport analysis of single-cell transcriptomics directs hypotheses prioritization and validation

The explosive growth of regulatory hypotheses from single-cell datasets demands accurate prioritization of hypotheses for in vivo validation. However, current computational methods emphasize overall accuracy in regulatory network reconstruction rather than prioritizing a limited set of causal transcription factors (TFs) that can be feasibly tested. We developed Haystack, a hybrid computational-biological algorithm that combines active learning and the concept of optimal transport theory to nominate and validate high-confidence causal hypotheses. Our novel approach efficiently identifies and prioritizes transient but causally-active TFs in cell lineages. We applied Haystack to single-cell observations, guiding efficient and cost-effective in vivo validations that reveal causal mechanisms of cell differentiation in Drosophila gut and blood lineages. Notably, all the TFs shortlisted for the final, imaging-based assays were validated as drivers of differentiation. Haystacks hypothesis-prioritization approach will be crucial for validating concrete discoveries from the increasingly vast collection of low-confidence hypotheses from single-cell transcriptomics.

bioinformatics↗

bioRxiv2022

Secure and federated linear mixed model association tests

Privacy-preserving algorithms for genome-wide association studies (GWAS) promise to facilitate data sharing across silos to accelerate new discoveries. However, existing approaches do not support an important, prevalent class of methods known as linear mixed model (LMM) association tests or would provide limited privacy protection, due to the high computational burden of LMMs under existing secure computation frameworks. Here we introduce SafeGENIE, an efficient and provably secure algorithm for LMM-based association studies, which allows multiple entities to securely share their data to jointly compute association statistics without leaking any intermediary results. We overcome the computational burden of LMMs by leveraging recent advances in LMMs and secure computation, as well as a novel scalable dimensionality reduction technique. Our results show that SafeGENIE obtains accurate association test results comparable to a state-of-the-art centralized algorithm (REGENIE), and achieves practical runtimes even for large datasets of up to 100K individuals. Our work unlocks the promise of secure and distributed algorithms for collaborative genomic studies.1

bioinformatics↗

bioRxiv2022

Identification and characterization of the T cell receptor (TCR) repertoire of the Cynomolgus macaque (Macaca Fascicularis)

BackgroundNon-human primates (NHP) are desirable as animal models of human disease because they share behavioral, physiological, and genomic traits with people. Hence, NHP recapitulate manifestations of disease not observed in other animal species. The Macaca fascicularis (i.e., Cynomolgus macaque) is an NHP species extensively used for biomedical research, but the TCR repertoire hasnt been characterized yet. ResultWe used the genomic sequences to design primers to identify the expressed TCR repertoire by single cell RNAseq. The data analysis from 22 unique samples were used to assign a functional status to each TCR genes. We identified and analyzed the TRA/D, TRB and TRG loci of the Cynomolgus macaque. ConclusionThe genomic organization of the Cynomolgus macaque has great similarity with Macaca mulatta (i.e., Rhesus macaque) and they shared >90% sequence similarity with the human TCR repertoire. These data will facilitate the analysis of T cell immunity in Cynomolgus macaques.

immunology↗

bioRxiv2021

Intrinsic instability of the dysbiotic microbiome revealed through dynamical systems inference at scale

Dynamical systems models are a powerful tool for analyzing interactions, stability, resilience, and other key properties in biomedically important microbial ecosystems, such as the gut microbiome. Challenges to modeling and inference in this setting include the large number of species present, and data sparsity/noise characteristics. Here, we introduce a Bayesian statistical method, the Microbial Dynamical Systems Inference Engine 2 (MDSINE2), which infers compact and interpretable ecosystems-scale dynamical systems models from microbiome time-series data. We model microbial dynamics as stochastic processes driven by inferred interaction modules, or groups of microbes with similar interaction structure and responses to perturbations. Additionally, we model the noise characteristics of sequencing and qPCR measurements to provide uncertainty quantification for all outputs. To evaluate MDSINE2, and provide a benchmarking resource for the community, we generated the most densely sampled microbiome time-series to date, which consists of a cohort of mice that received fecal transplants from a human donor and were then subjected to dietary and antibiotic perturbations. Benchmarking on simulated and real data demonstrate that MDSINE2 significantly outperforms state-of-the-art methods, and moreover identifies interaction modules that shed new light on ecosystems-scale interactions in the gut microbiome. We provide MDSINE2 as an open-source Python package at: https://github.com/gerberlab/MDSINE2.

systems biology↗

bioRxiv2021

Cellular and transcriptional diversity over the course of human lactation

Human breast milk is a dynamic fluid that contains millions of cells, but their identities and phenotypic properties are poorly understood. We used single-cell RNA-seq (scRNA-seq) to characterize the transcriptomes of cells from human breast milk (hBM) across lactational time from 3 to 632 days postpartum in 15 donors. We find that the majority of cells in human breast milk are lactocytes, a specialized epithelial subset, and cell type frequencies shift over the course of lactation yielding greater epithelial diversity at later points. Analysis of lactocytes reveals a continuum of cell states characterized by transcriptional changes in hormone, growth factor, and milk production related pathways. Generalized additive models suggest that one sub-cluster, LALBAlow epithelial cells, increase as a function of time postpartum, daycare attendance, and the use of hormonal birth control. We identify several sub-clusters of macrophages in hBM that are enriched for tolerogenic functions, possibly playing a role in protecting the mammary gland during lactation. Our description of the cellular components of breast milk, their association with maternal-infant dyad metadata and quantification of alterations at the gene and pathways levels provides the first detailed longitudinal picture of human breast milk cells across lactational time. This work paves the way for future investigations of how a potential division of cellular labor and differential hormone regulation might be leveraged therapeutically to support healthy lactation and potentially aid in milk production.

genomics↗

bioRxiv2021

HIV-specific CD8+ T-cells in tonsils express exhaustive TRM-like signatures

Lymphoid tissues are an important HIV reservoir site that persists in the face of antiretroviral therapy and natural immunity. Targeting these reservoirs by harnessing the antiviral activity of local tissue resident memory ( TRM) CD8+ T-cells is of great interest, but limited data exist on TRMs within lymph nodes of people living with HIV (PLWH). Here, we studied tonsil CD8+ T-cells obtained from PLWH and uninfected controls from South Africa. We show that these cells are preferentially located outside the germinal centers (GCs), the main reservoir site for HIV, and display a low cytolytic and transcriptionally TRM-like profile that is distinct from blood. In PLWH, CD8+ TRM-like cells are highly expanded and adopt a more cytolytic, activated and exhausted phenotype characterized by increased expression of CD69, PD-1 and perforin, but reduced CD127. This phenotype was enhanced in HIV-specific CD8+ T-cells from tonsils compared to matched blood. Single-cell profiling of these cells revealed a clear transcriptional signature of T-cell activation, clonal expansion and exhaustion ex-vivo. In contrast, this signature was absent from HIV-specific CD8+ T-cells in tonsils isolated from a natural HIV controller, who expressed lower levels of cell surface PD-1 and CXCR5, and reduced transcriptional evidence of T-cell activation, exhaustion and cytolytic activity. Thus, we show that HIV-specific TRM-like CD8+ T-cells in tonsils from non-HIV controllers are enriched for activation and exhaustion profiles compared to those in blood, suggesting that lymphoid HIV-specific CD8+ TRM cells are potentially ideal candidates for immunotherapy to modulate their ability to targeting the HIV reservoirs.

immunology↗