Biology subjects

Li, D.

Publications and source records attributed to Li, D..

At least 217 records · Page 12Linked to original sources

bioRxiv2024

Single-cell analysis of the epigenome and 3D chromatin architecture in the human retina

Most genetic risk variants linked to ocular diseases are non-protein coding and presumably contribute to disease through dysregulation of gene expression, however, deeper understanding of their mechanisms of action has been impeded by an incomplete annotation of the transcriptional regulatory elements across different retinal cell types. To address this knowledge gap, we carried out single-cell multiomics assays to investigate gene expression, chromatin accessibility, DNA methylome and 3D chromatin architecture in human retina, macula, and retinal pigment epithelium (RPE)/choroid. We identified 420,824 unique candidate regulatory elements and characterized their chromatin states in 23 sub-classes of retinal cells. Comparative analysis of chromatin landscapes between human and mouse retina cells further revealed both evolutionarily conserved and divergent retinal gene-regulatory programs. Leveraging the rapid advancements in deep-learning techniques, we developed sequence-based predictors to interpret non-coding risk variants of retina diseases. Our study establishes retina-wide, single-cell transcriptome, epigenome, and 3D genome atlases, and provides a resource for studying the gene regulatory programs of the human retina and relevant diseases.

genomics↗

bioRxiv2024

Intrahepatic crystals from elevated dietary cholesterol are sufficient to stiffen the liver

Chronic lipid accumulation is a hallmark of metabolic dysfunction-associated steatotic liver disease (MASLD), and dyslipidemia is associated with disease progression and poorer patient outcomes. Tissue stiffening is an established fibrogenic cue, but how changes in lipid accumulation affect tissue mechanics is not fully understood. Here we show that cholesterol-containing lipid crystals stiffen steatotic liver tissue. We show that rats fed elevated dietary cholesterol develop both liquid and solid cholesterol-containing lipid crystals in the liver. While steatotic livers without crystals are softer than control livers, livers with lipid crystals show increased baseline stiffness and compression stiffening as well as increased progression to fibrosis. Lipid extracts from livers containing crystals stiffen fibrous tissue mimics, while depletion of cholesterol using methyl--cyclodextrin reduces both crystal abundance and tissue stiffness. Our results demonstrate in a rat model that a high cholesterol diet leads to formation of liquid and solid crystals and that cholesterol-containing crystals stiffen tissues. This work implicates lipid crystals arising from dyslipidemia as a key driver of MASLD progression. The presence of cholesterol crystals could lead to new diagnostic tools for progressive MASLD and could be a therapeutic target.

physiology↗

bioRxiv2024

SenSet, a novel human lung senescence cell gene signature, identifies cell-specific senescence mechanisms

Cellular senescence is a major hallmark of aging. Senescence is defined as an irreversible growth arrest observed when cells are exposed to a variety of stressors including DNA damage, oxidative stress, or nutrient deprivation. While senescence is a well-established driver of aging and age-related diseases, it is a highly heterogeneous process with significant variations across organisms, tissues, and cell types. The relatively low abundance of senescence in healthy aged tissues represents a major challenge to studying senescence in a given organ, including the human lung. To overcome this limitation, we developed a Positive-Unlabeled (PU) learning framework to generate a comprehensive senescence marker gene list in human lungs (termed SenSet) using the largest publicly available single-cell lung dataset, the Human Lung Cell Atlas (HLCA). We validated SenSet in a highly complex ex vivo human 3D lung tissue culture model subjected to the senescence inducers bleomycin, doxorubicin, or irradiation, and established its sensitivity and accuracy in characterizing senescence. Using SenSet, we identified and validated cell-type specific senescence signatures in distinct lung cell populations upon aging and environmental exposures. Our study presents the first comprehensive analysis of senescent cells in the healthy aging lung and uncovers cell-specific gene signatures of senescence, presenting fundamental implications for our understanding of major lung diseases, including cancer, fibrosis, chronic obstructive pulmonary disease, or asthma.

genomics↗

bioRxiv2024

A molecularly defined brain circuit module for regulating the panic-like defensive state

Panic is an episode of strong defensive state, characterized by intense fear and severe physical symptoms such as elevated cardiorespiratory activities. How the brain generates panic state remains poorly understood. Here, we developed a robot-based experimental paradigm to evoke panic-like defensive state in mice. When stimulated by the robot, the mice exhibited jumping escapes and elevated cardiorespiratory activities. With this paradigm, we identified Cbln2-expressing (Cbln2+) neurons in the posterior hypothalamic nucleus (PHN) as a key neuronal population essential for the induction of panic-like defensive state. Activation of Cbln2+ PHN neurons induced behavioral and physical symptoms of panic-like defensive state. These neurons were strongly activated by noxious mechanical stimuli and encode jumping escape vigor. They were synaptically innervated by anxiety-associated brain areas and provoked panic-like defensive state via their projection to the periaqueductal gray. Together, our results reveal a molecularly defined circuit module that regulates the panic-like defensive state in mice.

neuroscience↗

bioRxiv2024

Causal disentanglement for single-cell representations and controllable counterfactual generation

Conducting disentanglement learning on single-cell omics data offers a promising alternative to traditional black-box representation learning by separating the semantic concepts embedded in a biological process. We present CausCell, which incorporates the causal relationships among disentangled concepts within a diffusion model to perform disentanglement learning, with the aim of increasing the explainability, generalizability and controllability of single-cell data, including spatial and temporal omics data, relative to those of the existing black-box representation learning models. Two quantitative evaluation scenarios, i.e., disentanglement and reconstruction, are presented to conduct the first comprehensive single-cell disentanglement learning benchmark, which demonstrates that CausCell outperforms the state-of-the-art methods in both scenarios. Additionally, CausCell can implement controllable generation by intervening with the concepts of single-cell data when given a causal structure. It also has the potential to uncover biological insights by generating counterfactuals from small and noisy single-cell datasets.

bioinformatics↗

bioRxiv2024

Assessment of DNA methylation from a single genomic region of ELOV2 is sufficient to predict chronological age

Estimation of chronological age is particularly informative in forensic contexts. Assessment of DNA methylation status allows for the prediction of age, though the accuracy and ease of manipulation may vary across different models. In this study, we started with a carefully designed discovery cohort recruiting more elderly subjects than other age categories, to diminish the effect of epigenetic drifting. We analyzed DNA methylation from a single genomic region of ELOV2, which was sufficient to construct an age-prediction model comprising 15 CpG sites. This model is further validated by an independent cohort as well as a multi-center test using trace dried bloodstains. The nature of our analytical pipeline, when combined the assessment of a single genomic locus with high-throughput sequencing, can easily be scaled up with low cost. Taken together, we propose a new age-prediction model featuring accuracy, ease of manipulation, high-throughput, and low cost. This model can be readily applied in both classic and newly emergent forensic contexts that require age estimation.

genetics↗

bioRxiv2024

Accurate and General DNA Representations Emerge from Genome Foundation Models at Scale

Language models applied to protein sequences have become a panacea, enabling therapeutics development, materials engineering, and core biology research. Despite the successes of protein language models, genome language models remain nascent. Recent studies suggest the bottleneck is data volume or modeling context size, since long-range interactions are widely acknowledged but sparsely annotated. However, it may be the case that even short DNA sequences are modeled poorly by existing approaches, and current models are unable to represent the wide array of functions encoded by DNA. To study this, we develop AIDO.DNA, a pretrained module for DNA representation in an AI-driven Digital Organism [1]. AIDO.DNA is a seven billion parameter encoder-only transformer trained on 10.6 billion nucleotides from a dataset of 796 species. By scaling model size while maintaining a short context length of 4k nucleotides, AIDO.DNA shows substantial improvements across a breadth of supervised, generative, and zero-shot tasks relevant to functional genomics, synthetic biology, and drug development. Notably, AIDO.DNA outperforms prior encoder-only architectures without new data, suggesting that new scaling laws are needed to achieve computeoptimal DNA language models. Models and code are available through Model-Generator in https://github.com/genbio-ai/AIDO and on Hugging Face at https://huggingface.co/genbio-ai.

bioinformatics↗

bioRxiv2024

Scaling Dense Representations for Single Cell with Transcriptome-Scale Context

Developing a unified model of cellular systems is a canonical challenge in biology. Recently, a wealth of public single-cell RNA sequencing data as well as rapid scaling of self-supervised learning methods have provided new avenues to address this longstanding challenge. However, rapid parameter scaling has been essential to the success of large language models in text and images, while similar scaling has not been attempted with Transformer architectures for cellular modeling. To produce accurate, transferable, and biologically meaningful representations of cellular systems, we develop AIDO.Cell, a pretrained module for representing gene expression and cellular systems in an AI-driven Digital Organism [1]. AIDO.Cell contains a series of 3M, 10M, 100M, and 650M parameter encoder-only dense Transformer models pre-trained on 50 million human cells from diverse tissues using a read-depth-aware masked gene expression pretraining objective. Unlike previous models, AIDO.Cell is capable of handling the entire human transcriptome as input without truncation or sampling tricks, thus learning accurate and general representations of the human cells entire transcriptional context. This pretraining with a longer context was enabled through FlashAttention-2, mixed precision, and large-scale distributed systems training. AIDO.Cell (100M) achieves state-of-the-art results in tasks such as zero-shot clustering, cell-type classification, and perturbation modeling. Our findings reveal interesting loss scaling behaviors as we increase AIDO.Cells parameters from 3M to 650M, providing insights for future directions in single-cell modeling. Models and code are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

Mixture of Experts Enable Efficient and Effective Protein Understanding and Design

Proteins play a fundamental role in life. Understanding the language of proteins offers significant potential for gaining mechanistic insights into biological systems and introduces new avenues for treating diseases, enhancing agriculture, and safeguarding the environment. While large protein language models (PLMs) like ESM2-15B and xTrimoPGLM-100B have achieved remarkable performance in diverse protein understanding and design tasks, these models, being dense transformer models, pose challenges due to their computational inefficiency during training and deployment. In this work, we introduce AIDO.Protein, a pretrained module for protein representation in an AI-driven Digital Organism [1]. AIDO.Protein is also the first mixture-of-experts (MoE) model in the protein domain, with model size scales to 16 billion parameters. Leveraging a sparse MoE architecture with 8 experts within each transformer block and selectively activating 2 experts for each input token, our model is significantly more efficient in training and inference. Through pre-training on 1.2 trillion amino acids collected from UniRef90 and ColabfoldDB, our model achieves state-of-the-art results across most tasks in the xTrimoPGLM benchmark. Furthermore, on over 280 ProteinGym Deep Mutational Scanning (DMS) assays, our model achieves nearly 99% of the overall performance of the best MSA-based model and significantly outperforms the previously reported state-of-the-art models that do not utilize MSA. We also adapted this model for structure-conditioned protein sequence generation tasks and achieved new SOTA in this domain. These results indicate that AIDO.Protein can serve as a strong foundation model for protein understanding and design. Models and codes are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

A Large-Scale Foundation Model for RNA Function and Structure Prediction

Originally marginalized as an intermediate in the information flow from DNA to protein, RNA has become the star of modern biology, holding the key to precision therapeutics, genetic engineering, evolutionary origins, and our understanding of fundamental cellular processes. Yet RNA is as mysterious as it is prolific, serving as an information store, a messenger, and a catalyst, spanning many underchar-acterized functional and structural classes. Deciphering the language of RNA is important not only for a mechanistic understanding of its biological functions but also for accelerating drug design. Toward this goal, we introduce AIDO.RNA, a pre-trained module for RNA in an AI-driven Digital Organism [1]. AIDO.RNA contains a scale of 1.6 billion parameters, trained on 42 million non-coding RNA (ncRNA) sequences at single-nucleotide resolution, and it achieves state-of-the-art performance on a comprehensive set of tasks, including structure prediction, genetic regulation, molecular function across species, and RNA sequence design. AIDO.RNA after domain adaptation learns to model essential parts of protein translation that protein language models, which have received widespread attention in recent years, do not. More broadly, AIDO.RNA hints at the generality of biological sequence modeling and the ability to leverage the central dogma to improve many biomolecular representations. Models and code are available through ModelGenerator in https://github.com/genbio-ai/AIDO and on Hugging Face.

bioinformatics↗

bioRxiv2024

Structure-Based Design of Covalent Nanobody Binders for a Thermostable Green Fluorescence Protein

Green fluorescence protein (GFP) lights up almost every aspects of life sciences. An ultra thermostable GFP (TGP), engineered from a coral GFP, offers potential advantages over traditional jellyfish-derived GFP due to its high stability. However, owing to its later discovery, TGP lacks the extensive toolsets available for GFP, such as heavy chain-only antibody binders known as nanobodies. In this study, we have structurally characterized TGP in complex with Sb92, a synthetic nanobody identified from a previous in vitro screening, revealing its precise three-dimensional epitope. This structural insight, alongside the previously characterized Sb44-TGP complex, have allowed us to rationally design disulfide bonds between antigen and antibody for tighter interactions. Using biochemical analyses, we identify two bridged complexes (TGP A18C-Sb44 V100C, TGP E118C-Sb92 S57C), with the TGP-Sb92 disulfide pair showing high resistance to reducing agents. Our study expands the toolkit available for TGP and should encourage its wider applications.

biochemistry↗

bioRxiv2024

Green Elegance T2T: An upgraded telomere-to-telomere genome assembly and annotation of looseleaf lettuce (Lactuca sativa var. crispa)

Lettuce (Lactuca sativa L.), a globally important vegetable crop, is cultivated in various horticultural varieties. Looseleaf lettuce is one of the main varieties that are consumed. Here, we report a telomere-to-telomere (T2T) genome assembly of looseleaf lettuce (L. sativa var. crispa cv. Green Elegance). By combining the sequencing data of previously reported Green Elegance v1.2, the 42.20 Gb clean ultra-long reads produced by Oxford Nanopore Technology were used to close the gaps and upgrade the genome assembly. After filling 11 gaps in Green Elegance v1.2, the final gapless telomere-to-telomere genome assembly is 2.58 Gb in length with a contig N50 of 282.47 Mb, containing nine centromeres and 18 telomeres. The genome contained 41,375 protein-coding genes, of which 99.10% were functionally annotated. This T2T genome of looseleaf lettuce will be valuable for the identification of genetic variation and the advancement of lettuce breeding.

genomics↗

bioRxiv2024

Quantification and potential functional relevanceof binding cooperativity of adjacent transcriptionfactors on DNA

In eukaryotes, expression of a particular gene is regulated by a combination of transcription factors (TFs) bound on regulatory regions of the genomic DNA (promoters and enhancers). Recent advances in genomic sequencing technology have allowed measurements of TFs footprints and binding affinities on DNA at the single-molecule level, permitting the probing of binding cooperativity among adjacent TFs. This necessitates quantitative descriptions of TFs binding cooperativity and understanding of its potential functional relevance. In this study, we show that, thermodynamically, the binding affinities of two adjacent TFs can either increase together (positive cooperativity) or decrease together (negative cooperativity), rather than changing in opposite directions. Their binding cooperativity can be quantified by the{gamma} coefficient, which is independent of TF concentrations, and can be determined by single-molecule binding reads either from in vitro thermodynamic condition or cellular condition of non-equilibrium steady state (NESS). The functional relevance of positive cooperativity, which has been extensively discussed in the literature, is the sigmoidal binding curve around a TF concentration threshold (analogous to oxygen binding to hemoglobin), whereas the functional relevance of negative cooperativity is two-fold. First, mutual exclusion of the two TFs enables bidirectional gene switching similar to CI-Cro system in phage{lambda} . Second, under a non-equilibrium steady-state condition, in which TFs often exhibit intranuclear concentration fluctuations, negative binding cooperativity assures fast TF dissociation from DNA and hence rapid response for gene expression regulation. Significance StatementIn eukaryotes, multiple transcription factors (TFs) bind to regulatory regions of DNA to control gene expression. The binding affinity of one TF to a specific DNA site is influenced by the binding of another TF at an adjacent site, exhibiting either positive or negative cooperativity (the TFs either strengthen or weaken each others binding). Here we present metric for such cooperativity from experimental measurables. Functionally, positive cooperativity assures sensitive response above a threshold of TF concentrations, whereas functional roles for negative cooperativity might be two folds: First, mutual exclusion of TFs binding enables bidirectional gene switching. Second, as TF concentrations oscillate under non-equilibrium steady-state condition, negative binding cooperativity assures fast TF dissociation, hence rapid switching of transcription.

biophysics↗

bioRxiv2024

Engineered mitochondrial A-to-G editors with enhanced efficiency and targeting scope

Mitochondrial base editing is a powerful technology, but the current A-to-G mitochondrial DNA (mtDNA) base editors are reluctant to achieve high efficiency, which is critical for mtDNA engineering. Through laboratory evolution, TadA-8e variants were discovered with substantially increased activity and expanded targeting compatibility, especially for previously unfavorite sequence contexts, when applied to both nuclear and mitochondrial ABEs. Further engineering of the mtDNA editors (eTd-mtABEs) dramatically reduced both DNA and RNA off-targeting effects and enhanced strand-selective A-to-G editing with substitution of DddA to DNA nickases. Moreover, the eTd-mtABEs induced up to 145-fold editing frequencies compared to previous mtDNA adenine base editors in rat cells, and installed targeted mutations in all injected rat embryos with up to 74% efficiency in founders, resulting generation of SNHL and Leigh syndrome models with severe defects. It suggests that eTd-mtABEs are promising mtDNA engineering technology for basic research and for translational studies. HighlightsO_LIEngineered TadA variants enhance nuclear and mitochondrial ABE activity in all sequence contexts. C_LIO_LIAdditional mutations in eTd-mtABEs further minimize both DNA and RNA off-targeting effects. C_LIO_LIEngineered TadA variants are compatible to DNA nickase assisted strand-selective mtDNA editing. C_LIO_LIeTd-mtABEs enable highly efficient generation of rat mtDNA disease models with severe phenotypes. C_LI

bioengineering↗

bioRxiv2024

RTF1 mediates epigenetic control of Th17 cell differentiation via H2B monoubiquitination

A gene encoding the transcription factor RTF1 has been associated with an increased risk of ulcerative colitis (UC). In this study, we investigated its function in modulating T cells expressing interleukin-17A (Th17 cells), a cardinal cell type promoting intestinal inflammation. Our results indicate that Rtf1 deficiency disrupts the differentiation of Th17 cells, while leaving regulatory T cells (Treg) unaffected. Mechanistically, RTF1 facilitates histone H2B monoubiquitination (H2Bub1), which requires its histone modification domain (HMD), for supporting Th17 cell function. Impaired Th17 differentiation was also observed in cells lacking the H2Bub1 E3 ligase subunit RNF40, an enzyme known to physically interact with RTF1. Thus, our study underscores the essential role of RTF1 in H2Bub1-mediated epigenetic regulation of Th17 cell differentiation. Understanding this process will likely provide valuable insights into addressing Th17-associated inflammatory disorders. Key PointsO_LIRTF1 is crucial for Th17 cell differentiation via histone H2B monoubiquitination C_LIO_LIRTF1 deficiency impairs Th17 differentiation, similar to H2Bub1 ligase RNF40 deficiency C_LIO_LIRTF1 and H2Bub1 colocalize on genes regulating Th17 cells, indicating an epigenetic role C_LI

immunology↗

bioRxiv2024

Marker gene fishing for single-cell data with complex heterogeneity

In single-cell studies, cells can be characterized with multiple sources of heterogeneity such as cell type, developmental stage, cell cycle phase, activation state, and so on. In some studies, many nuisance sources of heterogeneity (SOH) are of no interest, but may confound the identification of the SOH of interest, and thus affect the accurate annotate the corresponding cell subpopulations. In this paper, we develop B-Lightning, a novel and robust method designed to identify marker genes and cell subpopulations correponding to a SOH (e.g., cell activation status), isolating it from other sources of heterogeneity (e.g., cell type, cell cycle phase). B-Lightning uses an iterative approach to enrich a small set of trustworthy marker genes to more reliable marker genes and boost the signals of the SOH of interest. Multiple numerical and experimental studies showed that B-Lightning outperforms existing methods in terms of sensitivity and robustness in identifying marker genes. Moreover, it increases the power to differentiate cell subpopulations of interest from other heterogeneous cohorts. B-Lightning successfully identified new senescence markers in ciliated cells from human idiopathic pulmonary fibrosis (IPF) lung tissues, new T cell memory and effector markers in the context of SARS-COV-2 infections, and their synchronized patterns which were previously neglected. This paper highlights B-Lightnings potential as a powerful tool for single-cell data analysis, particularly in complex data sets where sources of heterogeneity of interest are entangled with numerous nuisance factors.

bioinformatics↗

bioRxiv2024

Bidirectional control of neurovascular coupling by cortical somatostatin interneurons

Neurovascular coupling, linking neuronal activity to cerebral blood flow, is altered early in neurological disorders and underlies functional brain imaging. This process involves numerous cellular players. Among them inhibitory interneurons receive increasing attention, but how they control blood flow remains elusive. This study elucidates the mechanisms by which somatostatin interneurons bidirectionally control neurovascular coupling. Patch clamp recordings in, ex vivo, cortical slices from mice expressing channelrhodopsin-2 in somatostatin interneurons, revealed that these neurons are supralinearly activated at low-frequencies (< 5 Hz) and efficiently photostimulated at frequencies up to 20 Hz. Ex vivo vascular imaging showed that low-frequency (2 Hz) photostimulation triggered vasodilation whereas high-frequency (20 Hz) photostimulation induced vasoconstriction. Histochemistry revealed that subpopulations of cortical somatostatin interneurons expressed the neuronal nitric oxide synthase, and/or neuropeptide Y to a greater extent. Consistently, pharmacological investigations showed that vasodilation induced by low-frequency photostimulation involves nitric oxide release and activation of soluble guanylate cyclase. In contrast, the vasoconstriction induced at high-frequency photostimulation involves neuropeptide Y release and activation of the Y1 vascular receptor. These findings provide valuable insights into neurovascular coupling and help to understand the cellular mechanism underlying the functional brain imaging signals.

neuroscience↗

bioRxiv2024

scGenePT: Is language all you need for modeling single-cell perturbations?

Modeling single-cell perturbations is a crucial task in the field of single-cell biology. Predicting the effect of up or down gene regulation or drug treatment on the gene expression profile of a cell can open avenues in understanding biological mechanisms and potentially treating disease. Most foundation models for single-cell biology learn from scRNA-seq counts, using experimental data as a modality to generate gene representations. Similarly, the scientific literature holds a plethora of information that can be used in generating gene representations using a different modality - language - as the basis. In this work, we study the effect of using both language and experimental data in modeling genes for perturbation prediction. We show that textual representations of genes provide additive and complementary value to gene representations learned from experimental data alone in predicting perturbation outcomes for single-cell data. We find that textual representations alone are not as powerful as biologically learned gene representations, but can serve as useful prior information. We show that different types of scientific knowledge represented as language induce different types of prior knowledge. For example, in the datasets we study, subcellular location helps the most for predicting the effect of single-gene perturbations, and protein information helps the most for modeling perturbation effects of interactions of combinations of genes. We validate our findings by extending the popular scGPT model, a foundation model trained on scRNA-seq counts, to incorporate language embeddings at the gene level. We start with NCBI gene card and UniProt protein summaries from the genePT approach and add gene function annotations from the Gene Ontology (GO). We name our model "scGenePT", representing the combination of ideas from these two models. Our work sheds light on the value of integrating multiple sources of knowledge in modeling single-cell data, highlighting the effect of language in enhancing biological representations learned from experimental data.

bioinformatics↗