Biology subjects

Gerstein, M.

Publications and source records attributed to Gerstein, M..

At least 37 records · Page 2Linked to original sources

bioRxiv2024

Complex genetic variation in nearly complete human genomes

Diverse sets of complete human genomes are required to construct a pangenome reference and to understand the extent of complex structural variation. Here, we sequence 65 diverse human genomes and build 130 haplotype-resolved assemblies (130 Mbp median continuity), closing 92% of all previous assembly gaps1,2 and reaching telomere-to-telomere (T2T) status for 39% of the chromosomes. We highlight complete sequence continuity of complex loci, including the major histocompatibility complex (MHC), SMN1/SMN2, NBPF8, and AMY1/AMY2, and fully resolve 1,852 complex structural variants (SVs). In addition, we completely assemble and validate 1,246 human centromeres. We find up to 30-fold variation in -satellite high-order repeat (HOR) array length and characterize the pattern of mobile element insertions into -satellite HOR arrays. While most centromeres predict a single site of kinetochore attachment, epigenetic analysis suggests the presence of two hypomethylated regions for 7% of centromeres. Combining our data with the draft pangenome reference1 significantly enhances genotyping accuracy from short-read data, enabling whole-genome inference3 to a median quality value (QV) of 45. Using this approach, 26,115 SVs per sample are detected, substantially increasing the number of SVs now amenable to downstream disease association studies.

genomics↗

bioRxiv2024

A Discard-and-Restart MD algorithm for the sampling of realistic protein transition states and enhance structure-based drug discovery

We introduce a Discard-and-Restart molecular dynamics (MD) algorithm tailored for the sampling of realistic protein transition states. It aids computational structure-based drug discovery by reducing the simulation times to compute transition pathways by up to 2000x. The algorithm iteratively performs short MD simulations and measures their proximity to a target state via a collective variable (CV) loss, which can be defined in a flexible fashion, locally or globally. Using the loss, if the trajectory proceeds toward the target, the MD simulation continues. Otherwise, it is discarded and a new MD simulation is restarted, with new initial velocities randomly drawn from a Boltzmann distribution. The discard-and-restart algorithm demonstrates efficacy and atomistic accuracy in capturing the folding pathways in several contexts: (1) fast-folding small protein domains; (2) the folding intermediate of the prion protein PrP; and (3) the spontaneous partial unfolding of -Tubulin, a crucial event for microtubule severing. During each iteration of the algorithm, we are able to perform AI-based analysis of the transitory conformations to find binding pockets, which could potentially represent druggable sites. Overall, our algorithm enables systematic and computationally efficient exploration of conformational landscapes, enhancing the design of ligands targeting dynamic protein states.

biophysics↗

bioRxiv2024

Predicting Spatially Resolved Gene Expression via Tissue Morphology using Adaptive Spatial GNNs

MotivationSpatial transcriptomics technologies, which generate a spatial map of gene activity, can deepen the understanding of tissue architecture and its molecular underpinnings in health and disease. However, the high cost makes these technologies difficult to use in practice. Histological images co-registered with targeted tissues are more affordable and routinely generated in many research and clinical studies. Hence, predicting spatial gene expression from the morphological clues embedded in tissue histological images, provides a scalable alternative approach to decoding tissue complexity. ResultsHere, we present a graph neural network based framework to predict the spatial expression of highly expressed genes from tissue histological images. Extensive experiments on two separate breast cancer data cohorts demonstrate that our method improves the prediction performance compared to the state-of-the-art, and that our model can be used to better delineate spatial domains of biological interest. Availabilityhttps://github.com/song0309/asGNN/

bioinformatics↗

bioRxiv2024

Homomorphic Encryption: An Application to Polygenic Risk Scores

1BackgroundPolygenic risk scores (PRSs) have emerged as a powerful tool in precision medicine, enabling personalized risk assessments for complex diseases. However, the use of sensitive genomic data in PRS calculations raises concerns about privacy and security. Fully homomorphic encryption (FHE) offers a promising solution by allowing computations on encrypted data, preserving the privacy of both genomic information and PRS models. MethodsHere, we present an application of FHE for encrypted PRS calculations using a particular protocol (CKKS) within the Lattigo library. Our approach involves a three-party system: clients (clinicians handling sensitive genetic data), modelers developing a PRS (academics or companies), and evaluators (a local hospital running the models while maintaining data confidentiality). We demonstrate the feasibility and accuracy of our approach by applying it to synthetic datasets of various sizes and to a robust 110k-single-nucleotide polymorphism (SNP) model for schizophrenia. The complete codebase and a sample dataset are available at https://github.com/gersteinlab/HEPRS. ResultsThe difference between traditional plaintext and encrypted PRS calculation results is negligible: the R2 is 0.999 and the mean squared error is 2.27 x 10-6. Moreover, while the encrypted calculation is roughly 1,000 times slower than conventional non-encrypted ones (when considering only the core PRS calculation), the computation remains feasible on a single-CPU node. For example, processing [~]1,100 individuals with [~]110k SNPs took six minutes and [~]65 GB of memory on a laptop computer. In addition, we investigated the impact of the encryption parameters on the computational time and accuracy in detail, showing the expected slowdown with higher security settings. ConclusionOur approach showcases the applicability and feasibility of using FHE on real-world PRS models. With the pressing need for privacy-preserving solutions in the era of precision medicine, our work serves as a pilot application, offering a simple use case and providing a detailed comparison and evaluation in terms of accuracy, cost, and scalability.

bioinformatics↗

bioRxiv2024

Epitope-anchored contrastive transfer learning for paired CD8+ T cell receptor-antigen recognition

Understanding the mechanisms of T-cell antigen recognition that underpin adaptive immune responses is critical for the development of vaccines, immunotherapies, and treatments against autoimmune diseases. Despite extensive research efforts, the accurate identification of T cell receptor (TCR)-antigen binding pairs remains a significant challenge due to the vast diversity and cross-reactivity of TCRs. Here, we propose a deep-learning framework termed Epitope-anchored Contrastive Transfer Learning (EPACT) tailored to paired human CD8+ TCRs from single-cell sequencing data. Harnessing the pre-trained representations and the contrastive co-embedding space, EPACT demonstrates state-of-the-art model generalizability in predicting TCR binding specificity for unseen epitopes and distinct TCR repertoires, offering potential values for practical outcomes in real-world scenarios. The contrastive learning paradigm achieves highly precise predictions for immunodominant epitopes and facilitates interpretable analysis of epitope-specific T cells. The TCR binding strength predicted by EPACT aligns well with the surge in spike-specific immune responses targeting SARS-CoV-2 epitopes after vaccination. We further fine-tune EPACT on TCR-epitope structural data to decipher the residue-level interactions involved in T-cell antigen recognition. EPACT not only exhibits superior capabilities in quantifying inter-chain distance matrices and identifying contact residue pairs but also corroborates the presence of molecular mimicry across multiple tumor-associated antigens. Together, EPACT can serve as a useful AI approach with significant potential in practical applications and contribute toward the development of TCR-based diagnostics and immunotherapies.

bioinformatics↗

bioRxiv2024

Transcriptional Determinism and Stochasticity Contribute to the Complexity of Autism Associated SHANK Family Genes

Precision of transcription is critical because transcriptional dysregulation is disease causing. Traditional methods of transcriptional profiling are inadequate to elucidate the full spectrum of the transcriptome, particularly for longer and less abundant mRNAs. SHANK3 is one of the most common autism causative genes. Twenty-four Shank3 mutant animal lines have been developed for autism modeling. However, their preclinical validity has been questioned due to incomplete Shank3 transcript structure. We applied an integrative approach combining cDNA-capture and long-read sequencing to profile the SHANK3 transcriptome in human and mice. We unexpectedly discovered an extremely complex SHANK3 transcriptome. Specific SHANK3 transcripts were altered in Shank3 mutant mice and postmortem brains tissues from individuals with ASD. The enhanced SHANK3 transcriptome significantly improved the detection rate for potential deleterious variants from genomics studies of neuropsychiatric disorders. Our findings suggest the stochastic transcription of genome associated with SHANK family genes.

genetics↗

bioRxiv2024

Validation of Enhancer Regions in Primary Human Neural Progenitor Cells using Capture STARR-seq

Genome-wide association studies (GWAS) and expression analyses implicate noncoding regulatory regions as harboring risk factors for psychiatric disease, but functional characterization of these regions remains limited. We performed capture STARR-sequencing of over 78,000 candidate regions to identify active enhancers in primary human neural progenitor cells (phNPCs). We selected candidate regions by integrating data from NPCs, prefrontal cortex, developmental timepoints, and GWAS. Over 8,000 regions demonstrated enhancer activity in the phNPCs, and we linked these regions to over 2,200 predicted target genes. These genes are involved in neuronal and psychiatric disease-associated pathways, including dopaminergic synapse, axon guidance, and schizophrenia. We functionally validated a subset of these enhancers using mutation STARR-sequencing and CRISPR deletions, demonstrating the effects of genetic variation on enhancer activity and enhancer deletion on gene expression. Overall, we identified thousands of highly active enhancers and functionally validated a subset of these enhancers, improving our understanding of regulatory networks underlying brain function and disease.

neuroscience↗

bioRxiv2024

Spatially Exploring RNA Biology in Archival Formalin-Fixed Paraffin-Embedded Tissues

Spatial transcriptomics has emerged as a powerful tool for dissecting spatial cellular heterogeneity but as of today is largely limited to gene expression analysis. Yet, the life of RNA molecules is multifaceted and dynamic, requiring spatial profiling of different RNA species throughout the life cycle to delve into the intricate RNA biology in complex tissues. Human disease-relevant tissues are commonly preserved as formalin-fixed and paraffin-embedded (FFPE) blocks, representing an important resource for human tissue specimens. The capability to spatially explore RNA biology in FFPE tissues holds transformative potential for human biology research and clinical histopathology. Here, we present Patho-DBiT combining in situ polyadenylation and deterministic barcoding for spatial full coverage transcriptome sequencing, tailored for probing the diverse landscape of RNA species even in clinically archived FFPE samples. It permits spatial co-profiling of gene expression and RNA processing, unveiling region-specific splicing isoforms, and high-sensitivity transcriptomic mapping of clinical tumor FFPE tissues stored for five years. Furthermore, genome-wide single nucleotide RNA variants can be captured to distinguish different malignant clones from non-malignant cells in human lymphomas. Patho-DBiT also maps microRNA-mRNA regulatory networks and RNA splicing dynamics, decoding their roles in spatial tumorigenesis trajectory. High resolution Patho-DBiT at the cellular level reveals a spatial neighborhood and traces the spatiotemporal kinetics driving tumor progression. Patho-DBiT stands poised as a valuable platform to unravel rich RNA biology in FFPE tissues to study human tissue biology and aid in clinical pathology evaluation.

systems biology↗

bioRxiv2024

A Variational Graph Partitioning Approach to Modeling Protein Liquid-liquid Phase Separation

Graph Neural Network (GNN)s have emerged as a powerful general-purpose tool for representation learning across many domains. Their efficacy often depends on having an optimal underlying graph for prediction. In many cases, the most relevant information comes from specific subgraphs. In this work, we introduce a novel GNN architecture, called Graph Partitioned GNN (GP-GNN), designed to partition graphs during the prediction process, thereby focusing attention on subgraphs containing the most relevant information. Our approach jointly learns task-dependent graph partitions and representations, making it particularly effective for predictive tasks where critical features may reside within initially unidentified subgraphs. Protein Liquid-Liquid Phase Separation (LLPS) is an important physical problem well-suited to our architecture, primarily because protein sub-domains called intrinsically disordered regions (IDR)s are known to play a crucial role in the phase separation process. LLPS plays an essential role in cellular processes and is known to be associated with various diseases (e.g ., neurodegenerative diseases and cancer). However, our ability to accurately predict which proteins undergo LLPS remains limited. In this study, we demonstrate how GP-GNN can be utilized to accurately predict LLPS by partitioning protein graphs into task-relevant subgraphs, such as those highlighting IDRs. Our model not only achieves state-of-the-art accuracy in predicting LLPS for both regulator and scaffold proteins but also offers valuable biological insights, that can be used to guide further downstream investigations. Notably, upon examining subgraphs identified by the GP-GNN, we show these are consistent with annotated IDRs.

bioinformatics↗

bioRxiv2024

Binding profiles for 954 Drosophila and C. elegans transcription factors revealtissue specific regulatory relationships

A catalog of transcription factor (TF) binding sites in the genome is critical for deciphering regulatory relationships. Here we present the culmination of the modERN (model organism Encyclopedia of Regulatory Networks) consortium that systematically assayed TF binding events in vivo in two major model organisms, Drosophila melanogaster (fly) and Caenorhabditis elegans (worm). We describe key features of these datasets, comprising 604 TFs identifying 3.6M sites in the fly and 350 TFs identifying 0.9 M sites in the worm. Applying a machine learning model to these data identifies sets of TFs with a prominent role in promoting target gene expression in specific cell types. TF binding data are available through the ENCODE Data Coordinating Center and at https://epic.gs.washington.edu/modERNresource, which provides access to processed and summary data, as well as widgets to probe cell type-specific TF-target relationships. These data are a rich resource that should fuel investigations into TF function during development.

genomics↗

bioRxiv2023

chronODE: A framework to integrate time-series multi-omics data based on ordinary differential equations combined with machine learning

Most functional genomic studies are conducted in steady-state conditions, therefore providing a description of molecular processes at a particular moment of cell differentiation or organismal development. Longitudinal studies can offer a deeper understanding of the kinetics underlying epigenetic events and their contribution to defining cell-type-specific transcriptional programs. Here we develop chronODE, a mathematical framework based on ordinary differential equations that uniformly models the kinetics of temporal changes in gene expression and chromatin features. chronODE employs biologically interpretable parameters that capture tissue-specific kinetics of genes and regulatory elements. We further integrate this framework with a neural-network architecture that can link and predict changes across different data modalities by solving multivariate time-series regressions. Next, we apply this framework to investigate region-specific kinetics of epigenome rewiring in the developing mouse brain, and we demonstrate that changes in chromatin accessibility within regulatory elements can accurately predict changes in the expression of putative target genes over the same time period. Finally, by integrating single-cell ATAC-seq data generated during the same time course, we show that regulatory elements characterized by fast activation kinetics in bulk measurements are active in early-appearing cell types, such as radial glial and other neural progenitors, whereas elements characterized by slow activation kinetics are specific to more differentiated cell types that emerge at later stages of brain development.

bioinformatics↗

bioRxiv2023

MolLM: Integrating 3D and 2D Molecular Representations with Biomedical Text via a Unified Pre-trained Language Model

MotivationThe current paradigm of deep learning models for the joint representation of molecules and text primarily relies on 1D or 2D molecular formats, neglecting significant 3D structural information that offers valuable physical insight. This narrow focus inhibits the models versatility and adaptability across a wide range of modalities. Conversely, the limited research focusing on explicit 3D representation tends to overlook textual data within the biomedical domain. ResultsWe present a unified pre-trained language model, MolLM, that concurrently captures 2D and 3D molecular information alongside biomedical text. MolLM consists of a text Transformer encoder and a molecular Transformer encoder, designed to encode both 2D and 3D molecular structures. To support MolLMs self-supervised pre-training, we constructed 160K molecule-text pairings. Employing contrastive learning as a supervisory signal for cross-modal information learning, MolLM demonstrates robust molecular representation capabilities across 4 downstream tasks, including cross-modality molecule and text matching, property prediction, captioning, and text-prompted molecular editing. Through ablation, we demonstrate that the inclusion of explicit 3D representations improves performance in these downstream tasks. Availability and implementationOur code, data, and pre-trained model weights are all available at https://github.com/gersteinlab/MolLM.

bioinformatics↗

bioRxiv2022

Revisiting the bad luck hypothesis: Cancer risk and aging are linked to replication-driven changes to the epigenome

Aging is the leading risk factor for cancer. While its been proposed that the age-related accumulation of somatic mutations drives this relationship, it is likely not the full story. Here, we show that both aging and cancer share a common epigenetic replication signature, which we modeled from DNA methylation data in extensively passaged immortalized human cells in vitro and tested on clinical tissues. This epigenetic signature of replication - termed CellDRIFT - increased with age across multiple tissues, distinguished tumor from normal tissue, and was escalated in normal breast tissue from cancer patients. Additionally, within-person tissue differences were correlated with both predicted lifetime tissue-specific stem cell divisions and tissue-specific cancer risk. Overall, our findings suggest that age-related replication drives epigenetic changes in cells, pushing them towards a more tumorigenic state. One sentence summaryCellular replication leaves an epigenetic fingerprint that may partially underly the age-associated increase in cancer risk.

cancer biology↗

bioRxiv2022

FAVOR: Functional Annotation of Variants Online Resource and Annotator for Variation across the Human Genome

Large-scale whole genome sequencing (WGS) studies and biobanks are rapidly generating a multitude of coding and non-coding variants. They provide an unprecedented resource for illuminating the genetic basis of human diseases. Variant functional annotations play a critical role in WGS analysis, result interpretation, and prioritization of disease- or trait-associated causal variants. Existing functional annotation databases have limited scope to perform online queries or are unable to functionally annotate the genotype data of large WGS studies and biobanks for downstream analysis. We develop the Functional Annotation of Variants Online Resources (FAVOR) to meet these pressing needs. FAVOR provides a comprehensive online multi-faceted portal with summarization and visualization of all possible 9 billion single nucleotide variants (SNVs) across the genome, and allows for rapid variant-, gene-, and region-level online queries. It integrates variant functional information from multiple sources to describe the functional characteristics of variants and facilitates prioritizing plausible causal variants influencing human phenotypes. Furthermore, a scalable annotation tool, FAVORannotator, is provided for functionally annotating and efficiently storing the genotype and variant functional annotation data of a large-scale sequencing study in an annotated GDS file format to facilitate downstream analysis. FAVOR and FAVORannotator are available at https://favor.genohub.org.

genetics↗

bioRxiv2022

A genome-wide atlas of recurrent repeat expansions in human cancer

Expansion of a single repetitive DNA sequence, termed a tandem repeat (TR), is known to cause more than 50 diseases. However, repeat expansions are often not explored beyond neurological and neurodegenerative disorders. In some cancers, mutations accumulate in short tracts of TRs (STRs), a phenomenon termed microsatellite instability (MSI); however larger repeat expansions have not been systematically analyzed in cancer. Here, we identified TR expansions in 2,622 cancer genomes, spanning 29 cancer types. In 7 cancer types, we found 160 recurrent repeat expansions (rREs); most of these (155/160) were subtype specific. We found that rREs were non-uniformly distributed in the genome with an enrichment near candidate cis-regulatory elements, suggesting a role in gene regulation. One rRE located near a regulatory element in the first intron of UGT2B7 was detected in 34% of renal cell carcinoma samples and was validated by long-read DNA sequencing. Moreover, targeting cells harboring this rRE with a rationally designed, sequence-specific DNA binder led to a dose-dependent decrease in cell proliferation. Overall, our results demonstrate that rREs are an important but unexplored source of genetic variation in human cancers, and we provide a comprehensive catalog for further study.

genomics↗

bioRxiv2022

Predicting A/B compartments from histone modifications using deep learning

Genomes fold into organizational units in the 3D space that can influence critical biological functions. In particular, the organization of chromatin into A and B compartments segregates its active regions from inactive regions. Compartments, evident in Hi-C contact matrices, have been used to describe cell-type specific changes in the A/B organization. However, obtaining Hi-C data for all cell and tissue types of interest is prohibitively expensive, which has limited the widespread consideration of compartment status. We present a prediction tool called Compartment prediction using Recurrent Neural Network (CoRNN) that models the relationship between the compartmental organization of the genome and histone modification enrichment. Our model predicts A/B compartments, in a cross-cell type setting, with an average area under the ROC curve of 90.9%. Our cell type-specific compartment predictions show high overlap with known functional elements. We investigate our predictions by systematically removing combinations of histone marks and find that H3K27ac and H3K36me3 are the most predictive marks. We then perform a detailed analysis of loci where compartment status cannot be accurately predicted from these marks. These regions represent chromatin with ambiguous compartmental status, likely due to variations in status within the population of cells. These ambiguous loci also show highly variable compartmental status between biological replicates in the same GM12878 cell type. Finally, we demonstrate the generalizability of our model by predicting compartments in independent tissue samples. Our software and trained model are publicly available at https://github.com/rsinghlab/CoRNN.

bioinformatics↗

bioRxiv2022

Capped nascent RNA sequencing reveals novel therapy-responsive enhancers in prostate cancer

Mounting evidence suggests that enhancer RNA (eRNA) transcription start sites (TSSs) provide higher sensitivity and specificity for enhancer identification than histone modifications and chromatin accessibility. The extent to which changes in eRNA transcription correspond to changes in enhancer activity, however, remains unclear. Here, we used precision run-on and capped RNA sequencing (PRO-cap) to assess changes in enhancer activity in response to treatment with the androgen receptor signaling inhibitor, enzalutamide (ENZ). We identified 6,189 high-confidence candidate enhancers in the human prostate cancer cell line, LNCaP; 853 of which demonstrated significant changes in activity in response to drug treatment. Notably, we found that 67% and 54% of drug-responsive enhancers did not show similar changes in activity in previous studies that utilized ChIP-seq and ATAC-seq, respectively. Strikingly, 79% of regions with increased eRNA transcription showed no other biochemical alterations, implying that PRO-cap can capture a set of precise changes in enhancer activity that classical approaches lack the sensitivity to detect. We performed in vivo functional validations of candidate enhancers and found that CRISPRi targeting of PRO-cap-specific drug-responsive enhancers impaired ENZ regulation of downstream target genes, suggesting that changes in eRNA TSSs mark true biological changes in enhancer activity with high sensitivity. Our study highlights the utility of using PRO-cap as a complementary approach to canonical biochemical methods for detecting precise changes in enhancer activity and, in particular, for better understanding disease progression and responses to treatment.

genomics↗

bioRxiv2022

scDVF: Data-driven Single-cell Transcriptomic Deep Velocity Field Learning with Neural Ordinary Differential Equations

Recent advances in single-cell RNA sequencing technology have provided unprecedented opportunities to simultaneously measure the gene expression profile and transcriptional velocity of individual cells, enabling us to sample gene regulatory network dynamics along developmental trajectories. However, traditional methods have faced challenges in modeling gene expression dynamics within individual cells due to sparse, non-linear (e.g., obligate heterodimer transcription factors), and high-dimensional measurements. Here, we present DeepVelo, a neural-network-based ordinary differential equation model that can learn non-linear, high-dimensional single-cell transcriptome dynamics and describe continuous gene expression changes within individual cells across time. We applied DeepVelo to multiple published datasets from different technical platforms and demonstrated its utility to 1) formulate transcriptome dynamics on different timescales, 2) measure the instability of cell states, and 3) identify developmental driver genes upstream of a signaling cascade. Benchmarking against state-of-the-art methods shows that DeepVelo can improve velocity field representation accuracy by at least 50% in out-of-sample cells. Further, perturbation studies revealed that single-cell dynamical systems may exhibit properties similar to those of chaotic systems. In summary, DeepVelo allows for the data-driven discovery of differential equations that delineate single-cell transcriptome dynamics. TeaserEmbedding neural networks into ordinary differential equations to model gene expression changes within single cells across time.

bioinformatics↗