Biology subjects

Ma, A.

Publications and source records attributed to Ma, A..

At least 73 records · Page 4Linked to original sources

bioRxiv2021

Biological network inference from single-cell multi-omics data using heterogeneous graph transformer

We present DeepMAPS (Deep learning-based Multi-omics Analysis Platform for Single-cell data) for biological network inference from single-cell multi-omics (scMulti-omics). DeepMAPS includes both cells and genes in a heterogeneous graph to simultaneously infer cell-cell, cell-gene, and gene-gene relations. The multi-head attention mechanism in a graph transformer considers the heterogeneous relation among cells and genes within both local and global context, making DeepMAPS robust to data noise and scale. We benchmarked DeepMAPS on 18 scMulti-omics datasets for cell clustering and biological network inference, and the results showed that our method outperformed various existing tools. We further applied DeepMAPS on lung tumor leukocyte CITE-seq data and matched diffuse small lymphocytic lymphoma scRNA-seq and scATAC-seq data. In both cases, DeepMAPS showed competitive performance in cell clustering and predicted biologically meaningful cell-cell communication pathways based on the inferred gene networks. Note that we deployed a webserver using DeepMAPS implementation equipped with multiple functions and visualizations to improve the feasibility and reproducibility of scMulti-omics data analysis. Overall, DeepMAPS represents a heterogeneous graph transformer for single-cell study and may benefit the use of scMulti-omics data in various biological systems.

bioinformatics↗

bioRxiv2021

A Mechanism for the Rare Fluctuation that Powers Protein Conformational Change

Most functional processes of biomolecules are rare events. Key to a rare event is the rare fluctuation that enables the energy activation process, which powers the system across the activation barrier. But the physical nature of this rare fluctuation and how it enables barrier crossing are unknown. With the help of a novel metric, the reaction capacity pC, that rigorously defines the beginning and parameterizes the progress of energy activation, the rare fluctuation was identified as a special phase-space condition that is necessary and sufficient for initiating systematic energy flow from the non-reaction coordinates into the reaction coordinates. The energy activation of a prototype biomolecular isomerization reaction is dominated by kinetic energy transferring into and accumulating in the reaction coordinates, administered by inertial forces alone. The two major reaction coordinates move in precise synergy, with one acting as a gating mechanism on the other. This mechanism is enabled by the structural features of biomolecules and may the cause of their unique functions that are not possible in small molecules.

biophysics↗

bioRxiv2021

Deep Transfer Learning of Drug Sensitivity by Integrating Bulk and Single-cell RNA-seq data

Massively bulk RNA sequencing databases incorporating drug screening have opened up an avenue to inform the optimal clinical application of cancer drugs. Meanwhile, the growing single-cell RNA sequencing (scRNA-seq) data contributes to improving therapeutic effectiveness by studying the heterogeneity of drug responses for cancer cell subpopulations. There is a clear significance in developing computational biology approaches to predict and interpret cancer drug response in single cell data from clinical samples. Here, we introduce scDEAL, a deep transfer learning framework for cancer drug response prediction at single-cell level by integrating large-scale bulk cell line data. The true innovation of scDEAL is to translate cancer cell line drug responses into predicting clinical drug responses via learning relations of gene expressions and drug responses at bulk-level and transfer to predict drug responses in scRNA-seq. Another innovation is the integrated gradient feature interpretation to infer a comprehensive set of signature genes to reveal potential drug resistance mechanisms. We benchmarked scDEAL on six scRNA-seq datasets and indicate its model interpretability through these case studies. We believe that this work may help study cell reprogramming, drug selection, and repurposing for improving therapeutic efficacy.

bioinformatics↗

bioRxiv2021

Transcriptome-wide association study of coronary artery disease identifies novel susceptibility genes

Transcriptome-wide association studies (TWAS) explore genetic variants affecting gene expression for association with a trait. Here we studied coronary artery disease (CAD) using this approach by first determining genotype-regulated expression levels in nine CAD relevant tissues by EpiXcan in two genetics-of-gene-expression panels, the Stockholm-Tartu Atherosclerosis Reverse Network Engineering Task (STARNET) and the Genotype-Tissue Expression (GTEx). Based on these data we next imputed gene expression in respective nine tissues from individual level genotype data on 37,997 CAD cases and 42,854 controls for a subsequent gene-trait association analysis. Transcriptome-wide significant association (P < 3.85e-6) was observed for 114 genes, which by genetic means were differentially expressed predominately in arterial, liver, and fat tissues. Of these, 96 resided within previously identified GWAS risk loci and 18 were novel (CAND1, EGFLAM, EZR, FAM114A1, FOCAD, GAS8, HOMER3, KPTN, MGP, NLRC4, RGS19, SDCCAG3, STX4, TSPAN11, TXNRD3, UFL1, WASF1, and WWP2). Gene set analyses showed that TWAS genes were strongly enriched in CAD-related pathways and risk traits. Associations with CAD or related traits were also observed for damaging mutations in 67 of these TWAS genes (11 novel) in whole-exome sequencing data of UK Biobank. Association studies in human genotype data of UK Biobank and expression-trait association statistics of atherosclerosis mouse models suggested that newly identified genes predominantly affect lipid metabolism, a classic risk factor for CAD. Finally, CRISPR/Cas9-based gene knockdown of RGS19 and KPTN in a human hepatocyte cell line resulted in reduced secretion of APOB100 and lipids in the cell culture medium. Taken together, our TWAS approach was able to i) prioritize genes at known GWAS risk loci and ii) identify novel genes which are associated with CAD.

genetics↗

bioRxiv2021

RESEPT: tissue architecture inference and visualization from spatially resolved transcriptomics

Spatially resolved transcriptomics provides a new way to define spatial contexts and understand biological functions in complex diseases. Although some computational frameworks can characterize spatial context via various clustering methods, the detailed spatial architectures and functional zonation often cannot be revealed and localized due to the limited capacities of associating spatial information. We present RESEPT, a deep-learning framework for characterizing and visualizing tissue architecture from spatially resolved transcriptomics. Given inputs as gene expression or RNA velocity, RESEPT learns a three-dimensional embedding with a spatial retained graph neural network from the spatial transcriptomics. The embedding is then visualized by mapping as color channels in an RGB image and segmented with a supervised convolutional neural network model. Based on a benchmark of sixteen 10x Genomics Visium spatial transcriptomics datasets on the human cortex, RESEPT infers and visualizes the tissue architecture accurately. It is noteworthy that, for the in-house AD samples, RESEPT can localize cortex layers and cell types based on a pre-defined region-or cell-type-specific genes and furthermore provide critical insights into the identification of amyloid-beta plaques in Alzheimers disease. Interestingly, in a glioblastoma sample analysis, RESEPT distinguishes tumor-enriched, non-tumor, and regions of neuropil with infiltrating tumor cells in support of clinical and prognostic cancer applications.

bioinformatics↗

bioRxiv2021

Verticillium dahliae effector VDAL protects MYB6 from degradation by interacting with PUB25/26 E3 ligases for enhancing Verticillium wilt resistance in Arabidopsis

Verticillium wilt is a severe plant disease that causes massive losses in multiple crops. Increasing the plant resistance to Verticillium wilt is a critical challenge worldwide. Here, we report that the hemibiotrophic Verticillium dahliae (V. dahliae)-secreted Asp f2-like protein VDAL causes leaf wilting when applied to cotton leaves in vitro, but enhances the resistance to V. dahliae when overexpressed in Arabidopsis or cotton without affecting the plant growth and development. VDAL protein interacts with Arabidopsis E3 ligases PUB25 and PUB26 (PUBs) and is ubiquitinated by PUBs in vitro. However, VDAL is not degraded by PUB25 or PUB26 in planta. Besides, the pub25 pub26 double mutant shows higher resistance to V. dahliae than the wild type. PUBs interact with the transcription factor MYB6 in a yeast two-hybrid screen. MYB6 promotes plant resistance to Verticillium wilt while PUBs ubiquitinate MYB6 and mediate its degradation. VDAL competes with MYB6 for binding to PUBs, and the role of VDAL in increasing Verticillium wilt resistance depends on MYB6. Taken together, these results suggest that plants evolute a strategy to utilize the invaded effector protein VDAL to resist the V. dahliae infection without causing a hypersensitive response (HR), Alternatively, in order to take nutrients from host cells, hemibiotrophic pathogens may use some effectors to keep plant cells alive during its infection in order to take nutrients from host cells. This study provides the molecular mechanism for plants increasing disease resistance when overexpressing some effector proteins without inducing HR, and may promote searching for more genes from pathogenic fungi or bacteria to engineer plant disease resistance. One-sentence summaryEctopically expressed VDAL in Arabidopsis alleviates the degradation of a positive disease response factor MYB6 through its interaction with PUB25/26 E3 ligases.

plant biology↗

bioRxiv2021

Multiple nucleotide polymorphism DNA markers for the accurate evaluation of genetic variations

DNA markers are an essential tool for the detection and evaluation of genetic variations, a central theme in genetics and biology. Effective markers must be highly reproducible, polymorphic, accurate and efficient to profile. We developed multiple dispersed nucleotide polymorphism (MNP) DNA marker and an efficient MNP genotyping method called MNP-Seq. The MNP marker was 17.48% more polymorphic than the highly polymorphic marker of microsatellites on a collection of hybrid rice plants. When applied to genotype more than 80,000 individual MNP markers of diploid rice and polyploidy hybrid cotton varieties which were notoriously difficult to genotype accurately, MNP-Seq finished in two days and achieved accuracies of 99.999% and 99.988%, respectively. We adopted MNP-Seq to reveal the ubiquitous, albeit subtle and neglected, genetic heterogeneities in homonyms of Nipponbare rice, a popular model organism for plant biology. This result raised a question on the consistency of the published results using the model plant. We also used MNP-Seq to accurately and efficiently determine the identities of plant varieties, a key but difficult problem for the protection of plant intellectual property rights. While being applied to plants in the current study, the MNP marker and MNP-Seq are general and readily applicable to similar problems in animals and micro-organisms.

molecular biology↗

bioRxiv2020

Assessing deep learning algorithms in cis-regulatory motif finding based on genomic sequencing data

Cis-regulatory motif finding is a crucial step in the detection of gene regulatory mechanisms using genomic data. Deep learning (DL) models have been utilized to denovoly identify motifs, and have been proven to outperform traditional methods. By 2020, twenty DL models have been developed to identify DNA and RNA motifs with diverse framework designs and implementation styles. Hence, it is beneficial to systematically compare their performances, which can facilitate researchers in selecting the appropriate tools for their motif analyses. Here, we carried out an in-depth assessment of the 20 models utilizing 1,043 genomic sequencing datasets, including 690 ENCODE ChIP-Seq, 126 cancer ChIP-Seq, 172 single-cell cleavages under targets and release using a nuclease, and 55 RNA CLIP-Seq. Four metrics were designed and investigated, including the accuracy of motif finding, the performance of DNA/RNA sequence classification, algorithm scalability, and tool usability. The assessment results demonstrated the high complementarity of the existing models, and it was determined that the most suitable model should primarily depend on the data size and type as well as the model outputs. A webserver was developed to allow efficient access of the identified motifs and effective utilization of high-performing DL models.

bioinformatics↗

bioRxiv2020

Control of neurogenic competence in mammalian hypothalamic tanycytes.

Hypothalamic tanycytes, radial glial cells that share many features with neuronal progenitors, generate small numbers of neurons in the postnatal hypothalamus, but the identity of these neurons and the molecular mechanisms that control tanycyte-derived neurogenesis are unknown. In this study, we demonstrate that tanycyte-specific disruption of the NFI family of transcription factors (Nfia/b/x) robustly stimulates tanycyte proliferation and tanycyte-derived neurogenesis. Single-cell RNA- and ATAC-Seq analysis reveals that NFI factors repress Shh and Wnt signaling in tanycytes, and small molecule modulation of these pathways blocks proliferation and tanycyte-derived neurogenesis in Nfia/b/x-deficient mice. We show that Nfia/b/x-deficient tanycytes give rise to multiple mediobasal hypothalamic neuronal subtypes that can mature, integrate into hypothalamic synaptic circuitry, and selectively respond to changes in internal states. These findings identify molecular mechanisms that control tanycyte-derived neurogenesis, suggesting a new therapeutic approach to selectively remodel the hypothalamic neural circuitry that controls homeostatic physiological processes.

neuroscience↗

bioRxiv2020

Inductive Inference of Gene Regulatory Network Using Supervised and Semi-supervised Graph Neural Networks

Discovering gene regulatory relationships and reconstructing gene regulatory networks (GRN) based on gene expression data is a classical, long-standing computational challenge in bioinformatics. Computationally inferring a possible regulatory relationship between two genes can be formulated as a link prediction problem between two nodes in a graph. Graph neural network (GNN) provides an opportunity to construct GRN by integrating topological neighbor propagation through the whole gene network. We propose an end-to-end gene regulatory graph neural network (GRGNN) approach to reconstruct GRNs from scratch utilizing the gene expression data, in both a supervised and a semi-supervised framework. To get better inductive generalization capability, GRN inference is formulated as a graph classification problem, to distinguish whether a subgraph centered at two nodes contains the link between the two nodes. A linked pair between a transcription factor (TF) and a target gene, and their neighbors are labeled as a positive subgraph, while an unlinked TF and target gene pair and their neighbors are labeled as a negative subgraph. A GNN model is constructed with node features from both explicit gene expression and graph embedding. We demonstrate a noisy starting graph structure built from partial information, such as Pearsons correlation coefficient and mutual information can help guide the GRN inference through an appropriate ensemble technique. Furthermore, a semi-supervised scheme is implemented to increase the quality of the classifier. When compared with established methods, GRGNN achieved state-of-the-art performance on the DREAM5 GRN inference benchmarks. GRGNN is publicly available at https://github.com/juexinwang/GRGNN. HighlightsO_LIWe present a novel formulation of graph classification in inferring gene regulatory relationships from gene expression and graph embedding. C_LIO_LIOur method leverages a powerful framework, gene regulatory graph neural network (GRGNN), which is flexible and powerful to ensemble statistical powers from a number of heuristic skeletons. C_LIO_LIOur results show GRGRNN outperforms previous supervised and unsupervised methods inductively on benchmarks. C_LIO_LIGRGNN can be interpreted and explained following the biological network motif hypothesis in gene regulatory networks. C_LI

bioinformatics↗

bioRxiv2020

scGNN: a novel graph neural network framework for single-cell RNA-Seq analyses

Single-cell RNA-sequencing (scRNA-Seq) is widely used to reveal the heterogeneity and dynamics of tissues, organisms, and complex diseases, but its analyses still suffer from multiple grand challenges, including the sequencing sparsity and complex differential patterns in gene expression. We introduce the scGNN (single-cell graph neural network) to provide a hypothesis-free deep learning framework for scRNA-Seq analyses. This framework formulates and aggregates cell-cell relationships with graph neural networks and models heterogeneous gene expression patterns using a left-truncated mixture Gaussian model. scGNN integrates three iterative multi-modal autoencoders and outperforms existing tools for gene imputation and cell clustering on four benchmark scRNA-Seq datasets. In an Alzheimers disease study with 13,214 single nuclei from postmortem brain tissues, scGNN successfully illustrated disease-related neural development and the differential mechanism. scGNN provides an effective representation of gene expression and cell-cell relationships. It is also a novel and powerful framework that can be applied to scRNA-Seq analyses.

bioinformatics↗

bioRxiv2020

Prediction of protein-protein interactions based on elastic net and deep forest

Prediction of protein-protein interactions (PPIs) helps to grasp molecular roots of disease. However, web-lab experiments to predict PPIs are limited and costly. Using machine-learning-based frameworks can not only automatically identify PPIs, but also provide new ideas for drug research and development from a promising alternative. We present a novel deep-forest-based method for PPIs prediction. First, pseudo amino acid composition (PAAC), autocorrelation descriptor (Auto), multivariate mutual information (MMI), composition-transition-distribution (CTD), and amino acid composition PSSM (AAC-PSSM), and dipeptide composition PSSM (DPC-PSSM) are adopted to extract and construct the pattern of PPIs. Secondly, elastic net is utilized to optimize the initial feature vectors and boost the predictive performance. Finally, GcForest-PPI model based on deep forest is built up. Benchmark experiments reveal that the accuracy values of Saccharomyces cerevisiae and Helicobacter pylori are 95.44% and 89.26%. We also apply GcForest-PPI on independent test sets and CD9-core network, crossover network, and cancer-specific network. The evaluation shows that GcForest-PPI can boost the prediction accuracy, complement experiments and improve drug discovery. The datasets and code of GcForest-PPI could be downloaded at https://github.com/QUST-AIBBDRC/GcForest-PPI/.

bioinformatics↗

bioRxiv2019

Induction of pancreatic tumor-selective ferroptosis through modulation of cystine import

Pancreatic ductal adenocarcinoma (PDA) is the third-leading cause of cancer mortality in the US and is highly resistant to classical, targeted, and immune therapies. We show that human PDA cells are dependent on the provision of exogenous cystine to avert a catastrophic accumulation of lipid reactive oxygen species (ROS) that, left unchecked, leads to ferroptotic cell death, both in vitro and in vivo. Using a dual-recombinase genetically engineered model, we found that acute deletion of Slc7a11 led to tumor-selective ferroptosis, tumor stabilizations/regressions, and extended overall survival. The mechanism of ferroptosis induction in PDA cells required the concerted depletion of both glutathione and coenzyme A, highlighting a novel branch of ferroptosis-relevant metabolism. Finally, we found that cystine depletion in vivo using the pre-IND agent cyst(e)inase phenocopied Slc7a11 deletion, inducing tumor-selective ferroptosis and disease stabilizations/regressions in the well-validated KPC model of PDA.\n\nOne Sentence SummaryGenetic and pharmacological targeting of cystine import induces pancreatic cancer-selective ferroptosis in vivo.

cancer biology↗

bioRxiv2019

A method for rapid selection of randomly induced mutations in a gene of interest using CRISPR/Cas9 mediated activation of gene expression.

We have developed a CRISPR/Cas9 based method for isolating randomly induced recessive lethal mutations in a gene of interest (GOI) by selection within the F1 progeny of a single genetic cross. Our method takes advantage of the ability to overexpress a GOI using CRISPR/Cas9 mediated activation of gene expression. In essence, the screening strategy is based upon the idea that if overexpression of a wild type allele can generate a phenotype, then overexpression of a newly induced loss-of-function allele will lack this phenotype. As a proof-of-principle, we used this method to select EMS induced mutations of the Drosophila gene hindsight (hnt). From approximately 45,000 F1 progeny we recovered 8 new EMS induced loss-of-function hnt alleles that we characterized as an allelic series of hypomorphic mutations. This new method can, in theory, be used to recover randomly induced point mutants in a GOI and can be applied to any circumstance where CRISPR/Cas9 mediated activation of gene expression is associated with lethality or a visible phenotype.

genetics↗

bioRxiv2019

A quiescent state following mild sensory arousal in Caenorhabditis elegans is potentiated by stress

An animals behavioral and physiological response to stressors includes changes to its responses to stimuli. How such changes occur is not well understood. Here we describe a Caenorhabditis elegans quiescent behavior, post-response quiescence (PRQ), which is modulated by the C. elegans response to cellular stressors. Following an aversive mechanical or blue light stimulus, worms respond first by briefly moving, and then become more quiescent for a period lasting tens of seconds. PRQ occurs at low frequency in unstressed animals, but is more frequent in animals that have experienced cellular stress due to ultraviolet light exposure as well as in animals following overexpression of epidermal growth factor (EGF). PRQ requires the function of the carboxypeptidase EGL-21 and the calcium-activated protein for secretion (CAPS) UNC-31, suggesting it has a neuropeptidergic mechanism. Although PRQ requires the sleep-promoting neurons RIS and ALA, it is not accompanied by decreased arousability, and does not appear to be homeostatically regulated, suggesting that it is not a sleep state. PRQ represents a simple, tractable model for studying how neuromodulatory states like stress alter behavioral responses to stimuli.

neuroscience↗

bioRxiv2019

Targeting the methyltransferase SETD8 impairs tumor cell survival and overcomes drug resistance independently of p53 status in multiple myeloma

Multiple myeloma (MM) is a malignancy of plasma cells that largely remains incurable. The search for new therapeutic targets is therefore essential. Here we show that a higher expression of the lysine methyltransferase SETD8, which is responsible for histone H4K20 mono-methylation, is an adverse prognosis factor associated with a poor outcome in two cohorts of newly diagnosed patients. Remarkably, primary malignant plasma cells are particularly addicted to SETD8 activity. Indeed, pharmacological inhibition of this enzyme by the chemical compound UNC0379 demonstrated a significantly higher toxicity in MM cells compared to normal cells from the bone marrow microenvironment. Moreover, RNA sequencing and functional studies revealed that SETD8 inhibition induces a mature non-proliferating plasma cell signature and an activation of the p53 canonical pathway, which together leads to an impairment of myeloma cell proliferation and survival. However, UNC0379 treatment triggers a deadly level of replicative stress in p53 deficient MM cells, indicating that the cytotoxicity associated with SETD8 inhibition is independent of the p53 status. Consistent with this, the combination of UNC0379 with the conventional cytotoxic agent melphalan strongly enhances DNA damage and overcomes drug resistance in myeloma cells. Thus, targeting SETD8 could be of therapeutic interest to improve MM treatment in high-risk patients independently of the p53 status.

cancer biology↗

bioRxiv2018

The Dynamic Conformational Landscapes of the Protein Methyltransferase SETD8

Elucidating conformational heterogeneity of proteins is essential for understanding protein functions and developing exogenous ligands for chemical perturbation. While structural biology methods can provide atomic details of static protein structures, these approaches cannot in general resolve less populated, functionally relevant conformations and uncover conformational kinetics. Here we demonstrate a new paradigm for illuminating dynamic conformational landscapes of target proteins. SETD8 (Pr-SET7/SET8/KMT5A) is a biologically relevant protein lysine methyltransferase for in vivo monomethylation of histone H4 lysine 20 and nonhistone targets. Utilizing covalent chemical inhibitors and depleting native ligands to trap hidden high-energy conformational states, we obtained diverse novel X-ray structures of SETD8. These structures were used to seed massively distributed molecular simulations that generated six milliseconds of trajectory data of SETD8 in the presence or absence of its cofactor. We used an automated machine learning approach to reveal slow conformational motions and thus distinct conformational states of SETD8, and validated the resulting dynamic conformational landscapes with multiple biophysical methods. The resulting models provide unprecedented mechanistic insight into how protein dynamics plays a role in SAM binding and thus catalysis, and how this function can be modulated by diverse cancer-associated mutants. These findings set up the foundation for revealing enzymatic mechanisms and developing inhibitors in the context of conformational landscapes of target proteins.

biophysics↗

bioRxiv2018

QUBIC2: A novel biclustering algorithm for large-scale bulk RNA-sequencing and single-cell RNA-sequencing data analysis

The combination of biclustering and large-scale gene expression data holds a promising potential for inference of the condition specific functional pathways/networks. However, existing biclustering tools do not have satisfied performance on high-resolution RNA-sequencing (RNA-Seq) data, majorly due to the lack of (i) a consideration of high sparsity of RNA-Seq data, e.g., the massive zeros or lowly expressed genes in the data, especially for single-cell RNA-Seq (scRNA-Seq) data, and (ii) an understanding of the underlying transcriptional regulation signals of the observed gene expression values. Here we presented a novel biclustering algorithm namely QUBIC2, for the analysis of large-scale bulk RNA-Seq and scRNA-Seq data. Key novelties of the algorithm include (i) used a truncated model to handle the unreliable quantification of genes with low or moderate expression, (ii) adopted the mixture Gaussian distribution and an information-divergency objective function to capture shared transcriptional regulation signals among a set of genes, (iii) utilized a Core-Dual strategy to identify biclusters and optimize relevant parameters, and (iv) developed a size-based P-value framework to evaluate the statistical significances of all the identified biclusters. Our method validation on comprehensive data sets of bulk and single cell RNA-seq data suggests that QUBIC2 had superior performance in functional modules detection and cell type classification compared with the other five widely-used biclustering tools. In addition, the applications of temporal and spatial data demonstrated that QUBIC2 can derive meaningful biological information from scRNA-Seq data. The source code for QUBIC2 can be freely accessed at https://github.com/maqin2001/qubic2.

bioinformatics↗