Biology subjects

Hao, M.

Publications and source records attributed to Hao, M..

At least 37 records · Page 2Linked to original sources

bioRxiv2024

A Novel Efficient Algorithm for Common Variants Genotyping from Low-Coverage Sequencing Data

Low-coverage whole-genome sequencing (LC-WGS) combined with imputation represents a cost-effective genotyping strategy for genome-wide association studies (GWAS) in population genetics. In this study, the Limpute algorithm was developed specifically for genotyping from low-coverage sequencing data, it extracts variant information from low-coverage sequencing data by the novel virtual probes and subsequently performs imputation through cross-reference between samples. Compared to the currently dominant algorithm for low-coverage sequencing data, GLIMPSE2, Limpute achieved similar imputation performance within common variants (r2>0.87) while the GLIMPSE2 has a runtime approximately five times longer than that of the Limpute. Furthermore, to fully evaluate the accuracy of genotype imputation by Limpute, we utilized high-coverage whole-genome sequencing data (30x), microarray data, and high-coverage whole-exome sequencing data (30x) as validation sets respectively. The results demonstrated that Limpute has a good imputation performance for common variants using low-coverage sequencing data (1x: r2 > 0.87; 3x: r2 > 0.92; 5x: r2 > 0.93). In summary, we present a highly efficient, low-cost algorithm for genotyping from low-coverage sequencing data, offering substantial support for genetic research.

bioinformatics↗

bioRxiv2024

Learning multi-cellular representations of single-cell transcriptomics data enables characterization of patient-level disease states

Single-cell RNA-seq (scRNA-seq) has become a prominent tool for studying human biology and disease. The availability of massive scRNA-seq datasets and advanced machine learning techniques has recently driven the development of single-cell foundation models that provide informative and versatile cell representations based on expression profiles. However, to understand disease states, we need to consider entire tissue ecosystems, simultaneously considering many different interacting cells. Here, we tackle this challenge by generating patient-level representations derived from multi-cellular expression context measured with scRNA-seq of tissues. We develop PaSCient, a novel model that employs a multi-level representation learning paradigm and provides importance scores at the individual cell and gene levels for fine-grained analysis across multiple cell types and gene programs characteristic of a given disease. We apply PaSCient to learn a disease model across a large-scale scRNA-seq atlas of 24.3 million cells from over 5,000 patients. Comprehensive and rigorous benchmarking demonstrates the superiority of PaSCient in disease classification and its multiple downstream applications, including dimensionality reduction, gene/cell type prioritization, and patient subgroup discovery.

bioinformatics↗

bioRxiv2024

Performance of GPT-4 on the American College of Radiology In-Service Examination

ObjectivesNo study has evaluated the ability of ChatGPT-4 to answer image-rich diagnostic radiology board exam questions or assessed for model drift in GPT-4s image interpretation abilities. In our study we evaluate GPT-4s performance on the American College of Radiology (ACR) 2022 Diagnostic Radiology In-Training Examination (DXIT). MethodsQuestions were sequentially input into GPT-4 with a standardized prompt. Each answer was recorded and overall accuracy was calculated, as was logic-adjusted accuracy, and accuracy on image-based questions. This experiment was repeated several months later to assess for model drift. ResultsGPT-4 achieved 58.5% overall accuracy, lower than the PGY-3 average (61.9%) but higher than the PGY-2 average (52.8%). Adjusted accuracy was 52.8%. GPT-4 showed significantly higher (p = 0.012) confidence for correct answers (87.1%) compared to incorrect (84.0%). Performance on image-based questions was notably poorer (p < 0.001) at 45.4% compared to text-only questions (80.0%), with adjusted accuracy for image questions of 36.4%. When the questions were repeated, GPT-4 chose a different answer 25.5% of the time and there was a small but insignificant decrease in accuracy. DiscussionGPT-4 performed between PGY-2 and PGY-3 levels on the 2022 DXIT, but significantly poorer on image-based questions, and with large variability in answer choices across time points. This study underscores the potential and risks of using minimally-prompted general AI models in interpreting radiologic images as a diagnostic tool. Implementers of general AI radiology systems should exercise caution given the possibility of spurious yet confident responses.

bioinformatics↗

bioRxiv2024

scMulan: a multitask generative pre-trained language model for single-cell analysis

Cells can be viewed as complex stories written by coordinated expression of genes. The success of AI large language models (LLMs) in mastering the human language inspired us to develop a large AI model scMulan with 368 million parameters to generate cell transcriptomics with designated attributes by learning the cell language. We defined a unified c-sentence to incorporate cell transcriptomics and meta-attributes, and pre-trained scMulan on the equivalence of 100 million human cells. Experiments showed that scMulan can generate designated pseudo transcriptomics, predict missing attributes of cells, reconstruct unobserved cells along functional gradients, and can help to identify driving regulators of cell fates. The generated data passed tests of current tools and can reflect the underlying biology.

bioinformatics↗

bioRxiv2024

A call for a unified and multimodal definition of cellular identity in the enteric nervous system

The enteric nervous system (ENS) is a tantalizing frontier in neuroscience. With the recent emergence of single cell transcriptomic technologies, this rare and poorly understood tissue has begun to be better characterized in recent years. A precise functional mapping of enteric neuron diversity is critical for understanding ENS biology and enteric neuropathies. Nonetheless, this pursuit has faced considerable technical challenges. By leveraging different methods to compare available primary mouse and human ENS datasets, we underscore the urgent need for careful identity annotation, achieved through the harmonization and advancements of wet lab and computational techniques. We took different approaches including differential gene expression, module scoring, co-expression and correlation analysis, unbiased biological function hierarchical clustering, data integration and label transfer to compare and contrast functional annotations of several independently reported ENS datasets. These analyses highlight substantial discrepancies stemming from an overreliance on transcriptomics data without adequate validation in tissues. To achieve a comprehensive understanding of enteric neuron identity and their functional context, it is imperative to expand tissue sources and incorporate innovative technologies such as multiplexed imaging, electrophysiology, spatial transcriptomics, as well as comprehensive profiling of epigenome, proteome, and metabolome. Harnessing human pluripotent stem cell (hPSC) models provides unique opportunities for delineating lineage trees of the human ENS, and offers unparalleled advantages, including their scalability and compatibility with genetic manipulation and unbiased screens. We encourage a paradigm shift in our comprehension of cellular complexity and function in the ENS by calling for large-scale collaborative efforts and research investments.

neuroscience↗

bioRxiv2024

scDirect: key transcription factor identification for directing cell state transitions based on single-cell multi-omics data

Cell state transitions are complicated processes that occur in various life activities. Understanding and artificially manipulating them have been longstanding challenges. Substantial experiments reveal that the transitions could be directed by several key transcription factors (TFs). Here we present scDirect, a computational framework to identify key TFs based on single-cell RNA-seq and ATAC-seq data. scDirect models the TF identification task as a linear inverse problem, and solve it with gene regulatory networks enhanced by a graph attention network. Through a benchmarking on a single-cell human embryonic stem cell atlas, we demonstrate the robustness and superiority of scDirect against alternative analysis methods on TF identification. We apply scDirect on various datasets, and scDirect exhibits high capability in identifying key TFs in cell differentiation and somatic cell conversion. Furthermore, scDirect can efficiently identify TF combinations for cell reprogramming, many of which have been experimentally validated. We envision that scDirect can utilize rapidly increasing single-cell datasets to identify key TFs for directing cell state transitions and may become an effective tool to facilitate cell engineering and regenerative medicine.

bioinformatics↗

bioRxiv2024

Tumor relapse-free survival prognosis related consistency between cancer tissue and adjacent normal tissue in drug repurposing for solid tumor via connectivity map

Traditional drug discovery encounters challenges, including high costs, time-intensive processes, and inherent risks. Drug repurposing emerges as a compelling alternative strategy to identify new indications for investigational or approved drugs, circumventing these obstacles. Among the various drug repurposing methods, the Disease-specific Signature-based Connectivity Map (Cmap) approach is widely utilized. However, the commonly employed method for constructing disease-specific signatures, known as Differentially Expressed Genes (DEG), faces issues related to inconsistencies between dysregulated genes and the prognosis of genes in tumor tissue, as well as discrepancies in prognosis genes between tumor and normal tissues. In this study, we propose a novel approach, Prognosis Consistency Scoring (PCS), aimed at addressing these inconsistencies. PCS measures the consistency of gene prognosis between tumor and normal tissues by combining the Recurrence-Free Survival (RFS) prognosis power of genes in both contexts. Disease-specific signatures are then constructed based on PCS, and drug repurposing is performed using the Cmap and Lincs Unified Environment (CLUE). Validation of predicted drugs is conducted using data from DrugBank and ClinicalTrials databases. Our findings reveal that the aforementioned inconsistencies are pervasive. Compared to signatures based on DEGs, PCS-based signatures exhibit superior performance, identifying more drugs with higher prediction accuracy, as confirmed by DrugBank annotations. Notably, a significant proportion of predicted drugs without corresponding indications were subsequently validated in the ClinicalTrials database. Additionally, PCS-based signatures demonstrate elevated disease specificity and association with Drug Related Gene (DRG).

bioinformatics↗

bioRxiv2023

Large Scale Foundation Model on Single-cell Transcriptomics

Large-scale pretrained models have become foundation models leading to breakthroughs in natural language processing and related fields. Developing foundation models in life science for deciphering the "languages" of cells and facilitating biomedical research is promising yet challenging. We developed a large-scale pretrained model scFoundation with 100M parameters for this purpose. scFoundation was trained on over 50 million human single-cell transcriptomics data, which contain high-throughput observations on the complex molecular features in all known types of cells. scFoundation is currently the largest model in terms of the size of trainable parameters, dimensionality of genes and the number of cells used in the pre-training. Experiments showed that scFoundation can serve as a foundation model for single-cell transcriptomics and achieve state-of-the-art performances in a diverse array of downstream tasks, such as gene expression enhancement, tissue drug response prediction, single-cell drug response classification, and single-cell perturbation prediction.

bioinformatics↗

bioRxiv2023

Projectome-defined subtypes and modular intra-hypothalamic subnetworks of peptidergic neurons

The hypothalamus plays a vital role in coordinating essential neuroendocrine, autonomic, and somatomotor responses for survival and reproduction. While previous studies have explored population-level projections of hypothalamic neurons, the specific innervation patterns of individual hypothalamic axons remain unclear. To understand the organization of hypothalamic axon projections, we conducted a comprehensive reconstruction of single-cell projectomes from 7,180 mouse hypothalamic neurons expressing specific neuropeptides. Our analysis identified 31 distinct subtypes based on projectome-defined characteristics, with many exhibiting long-range axon collateral projections to multiple brain regions. Notably, these subtypes selectively targeted specific subdomains within downstream areas, either unilaterally or bilaterally. Furthermore, we observed that individual peptidergic neuronal types encompassed multiple projectome-defined subtypes, explaining their diverse functional roles. Additionally, by examining intra-hypothalamic axon projections, we uncovered six modular subnetworks characterized by enriched intramodular connections and distinct preferences for downstream targets. This modular organization of the intra-hypothalamic network likely contributes to the coordinated organization of hypothalamic outputs. In summary, our comprehensive projectome analysis reveals the organizational principles governing hypothalamic axon projections, providing a framework for understanding the neural circuit mechanisms underlying the diverse and coordinated functions of the hypothalamus.

neuroscience↗

bioRxiv2023

xTrimoGene: An Efficient and Scalable Representation Learner for Single-Cell RNA-Seq Data

AO_SCPLOWBSTRACTC_SCPLOWThe advances in high-throughput sequencing technology have led to significant progress in measuring gene expressions in single-cell level. The amount of publicly available single-cell RNA-seq (scRNA-seq) data is already surpassing 50M records for human with each record measuring 20,000 genes. This highlights the need for unsupervised representation learning to fully ingest these data, yet classical transformer architectures are prohibitive to train on such data in terms of both computation and memory. To address this challenge, we propose a novel asymmetric encoder-decoder transformer for scRNA-seq data, called xTrimoGene, which leverages the sparse characteristic of the data to scale up the pre-training. This scalable design of xTrimoGene reduces FLOPs by one to two orders of magnitude compared to classical transformers while maintaining high accuracy, enabling us to train the largest transformer models over the largest scRNA-seq dataset today. Our experiments also show that the performance of xTrimoGene improves as we increase the model sizes, and it also leads to SOTA performance over various downstream tasks, such as cell classification, perturb-seq effect prediction, and drug combination prediction.

bioinformatics↗

bioRxiv2022

The C-terminal domain of RNase H and the C-terminus amino acid residue regulate virus release and autoprocessing of a defective HIV-1 possessing M50I and V151I changes in integrase

Previously, we reported that an HIV-1 variant containing Met-to-Ile change at codon 50 and Val-to-Ile mutation at codon 151 of integrase (IN), HIV(IN:M50I/V151I), was an impaired virus. Despite the mutations being in IN, the virus release was significantly suppressed (p < 0.0001) and the initiation of autoprocessing was inhibited; the mechanism of the defect remains unknown. In the current study, we attempted to identify the critical domains or amino acid (aa) residue(s) that promote defects in HIV(IN:M50I/V151I), using a series of variants, including truncated or aa-substituted RNase H (RH) or IN. The results demonstrated that virus release and the initiation of autoprocessing were regulated by the C-terminal domains (CTDs) of RH and IN. Further studies illustrated that Asp at codon 109 of RH CTD and Asp at the C terminus of IN induces the defect. This result indicated that the CTD of RH and IN in GagPol and particular aa positions in RH and IN regulated the virus release and the initiation of autoprocessing, and these sites could be potential targets for the developing new therapies.

microbiology↗

bioRxiv2022

Learning Spatially-Aware Representations of Transcriptomic Data via Transfer Learning

Profiling spatial variations of cellular composition and transcriptomic characteristics is important for understanding the physiology and pathology of tissues in health or diseases. Spatial transcriptomics (ST) data are powerful for depicting spatial gene expression but the currently dominating high-throughput technology is yet not at single-cell resolution. On the other hand, single-cell RNA-sequencing (SC) data provide high-throughput transcriptomic information at the single-cell level but lack spatial information. Integrating these two types of data would be ideal for revealing transcriptomic landscapes at single-cell resolution. We developed the method STEM (SpaTially aware EMbedding) for this purpose. It uses deep transfer learning to encode both ST and SC data into a unified spatially aware embedding space, and then uses the embeddings to infer the SC-ST mapping and predict pseudo-spatial adjacency between cells in the SC data. Semi-simulation and real data experiments verified that the embeddings preserved the spatial information and eliminated technical biases between SC and ST data. Besides, we can use attribution analysis in STEM to reveal genes whose expressions dominate spatial information. We applied STEM to data of human squamous cell carcinoma and of hepatic lobule to uncover the spatial localization of rare cell types data and reveal cell-type-specific gene expression variation along a spatial axis. STEM is a powerful tool for mapping SC and ST data to build single-cell level spatial transcriptomic landscapes, and can provide mechanistic insights into the spatial heterogeneity and microenvironments of tissues.

bioinformatics↗

bioRxiv2022

Decoding functional cell-cell communication events by multi-view graph learning on spatial transcriptomics

Cell-cell communication events (CEs) are mediated by multiple ligand-receptor pairs. Usually only a particular subset of CEs directly works for a specific downstream response in a particular microenvironment. We name them as functional communication events (FCEs) of the target responses. Decoding the FCE-target gene relations is important for understanding the machanisms of many biological processes, but has been intractable due to the mixing of multiple factors and the lack of direct observations. We developed a method HoloNet for decoding FCEs using spatial transcriptomic data by integrating ligand-receptor pairs, cell-type spatial distribution and downstream gene expression into a deep learning model. We modeled CEs as a multiview network, developed an attention-based graph learning method to train the model for generating target gene expression with the CE networks, and decoded the FCEs for specific downstream genes by interpreting the trained model. We applied HoloNet on three Visium datasets of breast cancer or liver cancer. It revealed the communication landscapes in tumor microenvironments, and uncovered how various ligand-receptor signals and cell types affect specific biological processes. We also validated the stability of HoloNet in a Slideseq-v2 dataset. The experiments showed that HoloNet is a powerful tool on spatial transcriptomic data to help revealing specific cell-cell communications in a microenvironment that shape cellular phenotypes.

bioinformatics↗

bioRxiv2022

Neural signals encoded by specific frequencies of closed-loop electrical stimulation structurally and functionally reconstruct spinal sensorimotor circuits after spinal cord injury

Epidural electrical stimulation restores locomotion in animals and humans with spinal cord injury (SCI). However, the coding rules underlying electrical stimulation remain poorly understood, which has greatly limited the application of such electrical neuromodulation techniques in SCI clinical treatment. To elucidate the coding rules of electrical stimulation on spinal sensorimotor circuit reconstruction after complete SCI, we initially developed a spinal-muscle closed-loop stimulation protocol to mimic feedforward and feedback electrical signals in spinal sensorimotor circuits. Afterwards, using methods of sensorimotor function evaluation, neural circuit tracing and neural signal recording, we discovered a unique stimulus frequency of 10-20 Hz under closed-loop conditions was required for structural and functional reconstruction of spinal sensorimotor circuits. The single-cell transcriptome analysis of activated motoneurons characterized molecular networks involved in spinal sensorimotor circuit reconstruction. This study provides insights into neural signal decoding during spinal sensorimotor circuit reconstruction, and indicates a technological approach for the clinical treatment of SCI.

neuroscience↗

bioRxiv2021

A human ensemble cell atlas (hECA) enables in data cell sorting

Single-cell omics data can characterize multifaceted features of massive cells and bring significant insights to biomedical researches. The accumulation of single-cell data provides growing resources for constructing atlases for all cells of a human organ or the whole body. The true assembly of a cell atlas should be cell-centric rather than file-centric. We proposed a unified information framework enabling seamless cell-centric data assembly and developed a human Ensemble Cell Atlas (hECA) as an instance. hECA version 1.0 assembled scRNA-seq data across multiple studies into one orchestrated data repository. It contains 1,093,299 labeled cells and metadata from 116 published human single-cell studies, covering 38 human organs and 11 systems. We invented three methods of applications based on the cell-centric assembly: "In data" cell sorting enables targeted data retrieval in the full atlas with customizable logic expressions; The "quantitative portraiture" system provides a multi-view presentation of biological entities (organs, cell types, and genes) of multiple granularities; The customizable reference creation allows users to use the cell-centric assembly to generate references for their own cell type annotations. Case studies on agile construction of user-defined sub-atlases and "in data" investigation of CAR-T off-targets in multiple organs showed the great potential of cell-centric atlas assembly.

bioinformatics↗

bioRxiv2021

Enteric glial cells favour accumulation of anti-inflammatory macrophages during the resolution of muscularis inflammation

ObjectiveMonocyte-derived macrophages (M{varphi}s) are crucial regulators during muscularis inflammation. However, it is unclear which microenvironmental factors are responsible for monocyte recruitment and neurotrophic M{varphi} differentiation in this paradigm. Here, we investigate M{varphi} heterogeneity at different stages of muscularis inflammation and determine how environmental cues can attract and activate tissue protective M{varphi}s. DesignSingle cell RNA sequencing was performed on immune cells from the muscularis of wild-type and CCR2-/- mice at different timepoints after muscularis inflammation. CX3CR1GFP/+ and CX3CR1CreERT2 R26YFP mice were analyzed by flow cytometry and immunofluorescence. The transcriptome of enteric glial cells (EGCs) was investigated using PLPCreERT2 Rpl22HA mice. In addition, we assessed the effect of supernatant from neurosphere-derived EGCs on monocyte differentiation based on the expression of pro- and anti-inflammatory factors. ResultsMuscularis inflammation induced marked alterations in mononuclear phagocyte populations associated with a rapid infiltration of Ly6c+ monocytes that locally acquired unique transcriptional states. Trajectory inference analysis revealed two main pro-resolving M{varphi} subpopulations during the resolution of muscularis inflammation, i.e. Cd206+ MhcIIhi and Timp2+ MhcIIlo M{varphi}s, which were both derived from CCR2+ monocytes. Interestingly, we found that EGCs were able to sense damage to the muscularis to stimulate monocyte recruitment and differentiation towards pro-resolving M{varphi}s via CCL2 and CSF1, respectively. ConclusionOur study provides a comprehensive insight into pro-resolving M{varphi} differentiation and their regulators during muscularis inflammation. We deepened our understanding in the interaction between EGCs and M{varphi}s, thereby highlighting pro-resolving M{varphi} differentiation as a potential novel therapeutic strategy for the treatment of intestinal inflammation.

immunology↗

bioRxiv2021

Natural Occurring Non-Synonymous Single Nucleotide Polymorphisms in Integrase and RNase H Regulate Assembly and Autoprocessing of HIV-1

Recently, a genome-wide association study using plasma HIV RNA reported that 14 naturally occurring non-synonymous single nucleotide polymorphisms (SNPs) in HIV derived from anti-retrovirus naive patients were associated with virus load (VL). However, the impact of each mutation on viral fitness was not investigated. Here, we constructed a series of HIV variants encoding each SNP using site-directed mutagenesis and examined their replicative abilities and biological properties. An HIV variant containing Met-to-Ile change at codon 50 in integrase (HIV(IN:M50I)) was found an impaired virus. Despite the mutation being in integrase, a quantification assay demonstrated that the virus release was significantly suppressed (P<0.001). Transmission electron microscopy analyses revealed that the accumulation of abnormal shapes of buds on the plasma membrane and the released virus particles retained immature forms. Western blot analysis demonstrated a defect in autoprocessing of GagPol and Gag polyproteins in the HIV(IN:M50I) particles. Forster Resonance Energy Transfer (FRET) assay displayed that GagPol containing IN:M50I (GagPol(IN:M50I)) significantly increased the efficiency of homodimerization (P<0.05) and heterodimerization with Gag (P<0.001), compared to GagPol(WT). HIV replication assay using a series of variants of HIV(IN:M50I) elucidated that the C-terminus residues, Asn at codon 288, plays a key role in the defect and the impaired maturation and replication capability was rescued by two other VL-associated SNPs, Ser-to-Asn change at codon 17 in integrase or Asn-to-Ser change at codon 79 in RNase H. These data demonstrate that Gag and GagPol assembly, virus release and autoprocessing are not only regulated by integrase but also RNase H. ImportanceA nascent HIV-1 is noninfectious. To become an infectious virus, Gag and GagPol polyproteins in the particles need to be cleaved by mature HIV protease (PR). PR is initially translated as an inactive embedded enzyme in a GagPol polyprotein. The embedded PR in homodimerized GagPol polyproteins catalyzes a proteolytic reaction to release the mature PR. This excision step by a self-cleavage is called autoprocessing. Here, during the evaluation of roles of naturally emerging non-synonymous SNPs in HIV RNA, we found that autoprocessing is inhibited by Met-to-Ile change at codon 50 in integrase in GagPol which increases the efficiency of heterodimerization with Gag. This defect was recovered by co-existing of other SNPs: Ser-to-Asn change at codon 17 in integrase or Asn-to-Ser mutation at codon 79 in RNase H, suggesting that autoprocessing is regulated by not only integrase but also RNase H in GagPol polyprotein.

microbiology↗

bioRxiv2021

A Comparative Analysis of Novel Deep Learning Models and Ensemble Learning Models to Predict the Allergenicity of Food Allergens

Traditional food allergen identification mainly relies on in vivo and in vitro experiments, which often needs a long period and high cost. The artificial intelligence (AI)-driven rapid food allergen identification method has solved the above mentioned two drawbacks and is becoming an efficient auxiliary tool. Aiming to overcome the limitations of lower accuracy of traditional machine learning models in predicting the allergenicity of food proteins, this work proposed to introduce deep learning model - transformer with self-attention mechanism, ensemble learning models (representative as Light Gradient Boosting Machine (LightGBM) eXtreme Gradient Boosting (XGBoost)) to solve the problem. In order to highlight the superiority of the proposed novel method, the study also selected various commonly used machine learning models as the baseline classifiers. The results of 5-fold cross-validation showed that the AUC of the deep model was the highest (0.9578), which was better than the ensemble learning and baseline algorithms. But the deep model need to be pre-trained, and the training cost is the highest. By comparing the characteristics of the transformer model and boosting models, it can be analyzed that, each model has its own advantage, which provides novel clues and inspiration for the rapid prediction of food allergens in the future.

bioinformatics↗