ABSTRACT
INTRODUCTION: Immune checkpoint inhibitors (ICIs) are potent and precise therapies for various cancer types, significantly improving survival rates in patients who respond positively to them. However, only a minority of patients benefit from ICI treatments. OBJECTIVES: Identifying ICI responders before treatment could greatly conserve medical resources, minimize potential drug side effects, and expedite the search for alternative therapies. Our goal is to introduce a novel deep-learning method to predict ICI treatment responses in cancer patients. METHODS: The proposed deep-learning framework leverages graph neural network and biological pathway knowledge. We trained and tested our method using ICI-treated patients' data from several clinical trials covering melanoma, gastric cancer, and bladder cancer. RESULTS: Our results demonstrate that this predictive model outperforms current state-of-the-art methods and tumor microenvironment-based predictors. Additionally, the model quantifies the importance of pathways, pathway interactions, and genes in its predictions. A web server for IRnet has been developed and deployed, providing broad accessibility to users at https://irnet.missouri.edu. CONCLUSION: IRnet is a competitive tool for predicting patient responses to immunotherapy, specifically ICIs. Its interpretability also offers valuable insights into the mechanisms underlying ICI treatments.
ABSTRACT
Large protein language models (PLMs) present excellent potential to reshape protein research by encoding the amino acid sequences into mathematical and biological meaningful embeddings. However, the lack of crucial 3D structure information in most PLMs restricts the prediction capacity of PLMs in various applications, especially those heavily depending on 3D structures. To address this issue, we introduce S-PLM, a 3D structure-aware PLM utilizing multi-view contrastive learning to align the sequence and 3D structure of a protein in a coordinate space. S-PLM applies Swin-Transformer on AlphaFold-predicted protein structures to embed the structural information and fuses it into sequence-based embedding from ESM2. Additionally, we provide a library of lightweight tuning tools to adapt S-PLM for diverse protein property prediction tasks. Our results demonstrate S-PLM's superior performance over sequence-only PLMs, achieving competitiveness in protein function prediction compared to state-of-the-art methods employing both sequence and structure inputs.
ABSTRACT
Despite clinical use of immunosuppressive agents, the immunopathogenesis of minimal change disease (MCD) and focal segmental glomerulosclerosis (FSGS) remains unclear. Src homology 3-binding protein 2 (SH3BP2), a scaffold protein, forms an immune signaling complex (signalosome) with 17 other proteins, including phospholipase Cγ2 (PLCγ2) and Rho-guanine nucleotide exchange factor VAV2 (VAV2). Bioinformatic analysis of human glomerular transcriptome (Nephrotic Syndrome Study Network cohort) revealed upregulated SH3BP2 in MCD and FSGS. The SH3BP2 signalosome score and downstream MyD88, TRIF, and NFATc1 were significantly upregulated in MCD and FSGS. Immune pathway activation scores for Toll-like receptors, cytokine-cytokine receptor, and NOD-like receptors were increased in FSGS. Lower SH3BP2 signalosome score was associated with MCD, higher estimated glomerular filtration rate, and remission. Further work using Sh3bp2KI/KI transgenic mice with a gain-in-function mutation showed ~6-fold and ~25-fold increases in albuminuria at 4 and 12 weeks, respectively. Decreased serum albumin and unchanged serum creatinine were observed at 12 weeks. Sh3bp2KI/KI kidney morphology appeared normal except for increased mesangial cellularity and patchy foot process fusion without electron-dense deposits. SH3BP2 co-immunoprecipitated with PLCγ2 and VAV2 in human podocytes, underscoring the importance of SH3BP2 in immune activation. SH3BP2 and its binding partners may determine the immune activation pathways resulting in podocyte injury leading to loss of the glomerular filtration barrier.
Subject(s)
Glomerulosclerosis, Focal Segmental , Nephrosis, Lipoid , Nephrotic Syndrome , Animals , Humans , Mice , Adaptor Proteins, Signal Transducing/genetics , Adaptor Proteins, Signal Transducing/metabolism , Glomerulosclerosis, Focal Segmental/genetics , Glomerulosclerosis, Focal Segmental/metabolism , Kidney/pathology , Kidney Glomerulus/pathology , Mice, Transgenic , Nephrosis, Lipoid/pathology , Nephrotic Syndrome/metabolism , Phospholipase C gamma/genetics , Phospholipase C gamma/metabolismABSTRACT
Predicting protein localization and understanding its mechanisms are critical in biology and pathology. In this context, we propose a new web application of MULocDeep with improved performance, result interpretation, and visualization. By transferring the original model into species-specific models, MULocDeep achieved competitive prediction performance at the subcellular level against other state-of-the-art methods. It uniquely provides a comprehensive localization prediction at the suborganellar level. Besides prediction, our web service quantifies the contribution of single amino acids to localization for individual proteins; for a group of proteins, common motifs or potential targeting-related regions can be derived. Furthermore, the visualizations of targeting mechanism analyses can be downloaded for publication-ready figures. The MULocDeep web service is available at https://www.mu-loc.org/.
Subject(s)
Proteins , Software , Amino Acids/metabolism , Computational Biology/methods , Protein Transport , Proteins/chemistry , InternetABSTRACT
Single-cell multi-omics (scMulti-omics) allows the quantification of multiple modalities simultaneously to capture the intricacy of complex molecular mechanisms and cellular heterogeneity. Existing tools cannot effectively infer the active biological networks in diverse cell types and the response of these networks to external stimuli. Here we present DeepMAPS for biological network inference from scMulti-omics. It models scMulti-omics in a heterogeneous graph and learns relations among cells and genes within both local and global contexts in a robust manner using a multi-head graph transformer. Benchmarking results indicate DeepMAPS performs better than existing tools in cell clustering and biological network construction. It also showcases competitive capability in deriving cell-type-specific biological networks in lung tumor leukocyte CITE-seq data and matched diffuse small lymphocytic lymphoma scRNA-seq and scATAC-seq data. In addition, we deploy a DeepMAPS webserver equipped with multiple functionalities and visualizations to improve the usability and reproducibility of scMulti-omics data analysis.
Subject(s)
Benchmarking , Data Analysis , Reproducibility of Results , Cluster Analysis , Electric Power Supplies , Single-Cell AnalysisABSTRACT
As global climate change intensifies, people are paying increasing attention to the impact of temperature changes on adverse mental health outcomes, especially depression. While increasing attention has been paid to the effect of temperature, there is little research on the effect of humidity. We aimed to investigate the association between humidex, an index combining temperature and humidity to reflect perceived temperature, and outpatient visits for depression from 2014 to 2019 in Chongqing, the largest and one of the most hot and humid cities of China. We also aimed to further identify susceptible subgroups. A distributed lag non-linear model (DLNM) was used to explore the concentration-response relationship between humidex and depression outpatient visits. Hierarchical analysis was carried out by age and gender. A total of 155,436 visits for depression were collected from 2014 to 2019 (2191 days). We found that depression outpatient visits were significantly associated with extremely high humidex (≥40). The significant positive single-lag day effect existed at lag 0 (RR = 1.029, 95%CI: 1.000-1.059) to lag 2 (RR = 1.01, 95%CI: 1.004-1.028), and lag 12 (RR = 1.013, 95%CI: 1.002-1.024). The significant cumulative adverse effects lasted from lag 01 to lag 014. Hierarchical analyses showed that females and the elderly (≥60 years) appeared to be more susceptible to extremely high humidex. The attributable numbers (AN) and fraction (AF) of extremely high humidex on depression outpatients were 1709 and 1.10%, respectively. Extremely high humidex can potentially increase the risk of depression, especially in females and the elderly. More protective measures should be taken in vulnerable populations.
Subject(s)
Depression , Female , Humans , Aged , Time Factors , Temperature , Humidity , ChinaABSTRACT
BACKGROUND: Ambient air pollution has been linked to gestational complications. However, the evidence on the relationship between air pollution and fetal distress is limited. OBJECTIVES: To investigate the relationship between maternal short-term air pollution exposure and fetal distress, and to identify a potential susceptible population. METHODS: This matched case-control study, involving 313 pregnancy women with fetal distress was conducted in Xi'an, the largest city in Northwest China from 2013 to 2016. Each woman with fetal distress was randomly matched with four women without fetal distress of the same age, same gestational week, and registration in the same period (n = 1252). Inverse distance-weighted (IDW) interpolation was applied to estimate maternal air pollution exposure based on the residential addresses. We employed conditional logistic regression model to evaluate the relationship between air pollutants and fetal distress. Distributed lag nonlinear model (DLNM) was performed to examine the exposure-response relationship between air pollutants and fetal distress. RESULTS: Maternal short-term exposure to PM10, PM2.5-10 (PMc), SO2, NO2, and CO was associated with increased risk of fetal distress. Each 10 µg/m3 increment in PM10, PMc, SO2 at lag 014, and NO2 at lag 010, the odds ratio (ORs) of fetal distress were 1.027 (95 % confidence interval (CI): 1.004, 1.050), 1.058 (95 % CI: 1.014, 1.105), 1.140 (95 % CI: 1.029, 1.264), and 1.158 (95 % CI: 1.046, 1.283), respectively. Similarly, with a 0.1 mg/m3 increment in CO at lag 014, the OR of fetal distress was 1.029 (95 % CI: 1.002, 1.058). Stratified analyses showed that the estimate associations of PM10, PM2.5 and CO appeared to be stronger, although not statistically significantly, among women with gestational complications. CONCLUSION: Maternal short-term exposure to ambient air pollution may increase the risk of fetal distress. Understanding the detrimental role of air pollution in fetal distress can help us better develop preventative methods in reducing its' impact on maternal and fetal health.
Subject(s)
Air Pollutants , Air Pollution , Pregnancy , Humans , Female , Case-Control Studies , Nitrogen Dioxide , Fetal Distress/chemically induced , Environmental Exposure , Air Pollution/analysis , Air Pollutants/analysis , Maternal Exposure , China/epidemiology , Particulate Matter/analysisABSTRACT
Previous researches have reported the association between air pollution and various diseases. However, few researches have investigated whether air pollutants are associated with the economic loss resulting from patients' hospitalization, especially the economic loss of hospitalization due to acute cardiovascular events. The purpose of our research was to explore the association between the levels of carbon monoxide (CO), taken as an index of pollution, and the hospitalization costs of myocardial infarction (MI), and the potential effect modification by the ABO blood group. A total of 3237 MI inpatients were included in this study. A multiple linear regression model was used to evaluate the association between ambient CO levels and hospitalization costs of MI patients. Moreover, we performed stratified analyses by age, gender, body mass index (BMI), season, hypertension, and ABO blood types. There was a positive association between the levels of CO in the air and the costs of hospitalization caused by MI. Furthermore, such association was stronger in males, BMI ≥25, <65 years, with hypertension, and non-O blood group. Interestingly, we found the association was particularly significant in patients with blood group B. Overall, our study first found that ambient CO levels could have an impact on the hospitalization costs for MI patients, and those with blood group B can be more sensitive.
Subject(s)
Air Pollutants , Air Pollution , Hypertension , Myocardial Infarction , Male , Humans , Carbon Monoxide/analysis , ABO Blood-Group System/analysis , Air Pollution/analysis , Air Pollutants/toxicity , Air Pollutants/analysis , Hospitalization , Myocardial Infarction/epidemiology , Myocardial Infarction/chemically induced , Hypertension/chemically inducedABSTRACT
Evidence of the short-term effects of ambient sulfur dioxide (SO2) exposure on the economic burden of ischemic stroke is limited. This study aimed to explore the association between short-term ambient SO2 exposure and hospitalization costs for ischemic stroke in Chongqing, the most populous city in China. The hospital-based study included 7271 ischemic stroke inpatients. Multiple linear regression models were used to estimate the association between SO2 concentration and hospitalization costs. Propensity score matching was used to compare the patients' characteristics when exposed to SO2 concentrations above and below 20 µg/m3. It is found that short-term SO2 exposure was positively correlated with the hospitalization costs of ischemic stroke. The association was more evident in males, people younger than 65, and people hospitalized in the cool seasons. Besides, among the components of hospitalization costs, medicine costs were most significantly associated with SO2. More interesting, the lower concentration of SO2, the higher costs associated with 1 µg/m3 SO2 change. Above all, SO2 was positively associated with hospitalization costs of ischemic stroke, even at its low levels. The measures to reduce the level of SO2 can help reduce the burden of ischemic stroke.
Subject(s)
Air Pollutants , Air Pollution , Ischemic Stroke , Male , Humans , Air Pollutants/analysis , Sulfur Dioxide/analysis , Air Pollution/analysis , Particulate Matter/analysis , Environmental Exposure/analysis , Hospitalization , China , Hospitals , Nitrogen DioxideABSTRACT
Increased fluid-flow shear stress (FFSS) contributes to hyperfiltration-induced podocyte and glomerular injury resulting in progression of chronic kidney disease (CKD). We reported that increased FFSS in vitro and in vivo upregulates PGE2 receptor EP2 (but not EP4 expression), COX2-PGE2 -EP2 axis, and EP2-linked Akt-GSK3ß-ß-catenin signaling pathway in podocytes. To understand and use the disparities between PGE2 receptors, specific agonists, and antagonists of EP2 and EP4 were used to assess phosphorylation of Akt, GSK3ß and ß-catenin in podocytes using Western blotting, glomerular filtration barrier function using in vitro albumin permeability (Palb ) assay, and mitigation of hyperfiltration-induced injury in unilaterally nephrectomized (UNX) mice at 1 and 6 months. Results show an increase in Palb by PGE2 , EP2 agonist (EP2AGO ) and EP4 antagonist (EP4ANT ), but not by EP2 antagonist (EP2ANT ) or EP4 agonist (EP4AGO ). Pretreatment with EP2ANT blocked the effect of PGE2 or EP2AGO on Palb . Modulation of EP2 and EP4 also induced opposite effects on phosphorylation of Akt and ß-Catenin. Individual agonists or antagonists of EP2 or EP4 did not induce significant improvement in albuminuria in UNX mice. However, treatment with a combination EP2ANT + EP4AGO for 1 or 6 months caused a robust decrease in albuminuria. EP2ANT + EP4AGO combination did not impact adaptive hypertrophy or increased serum creatinine. Observed differences between expression of EP2 and EP4 on the glomerular barrier highlight these receptors as potential targets for intervention. Safe and effective mitigating effect of EP2ANT + EP4AGO presents a novel opportunity to delay the progression of hyperfiltration-associated CKD as seen in transplant donors.
Subject(s)
Receptors, Prostaglandin E, EP2 Subtype , Renal Insufficiency, Chronic , Albumins , Albuminuria , Animals , Creatinine , Cyclooxygenase 2 , Dinoprostone/metabolism , Glycogen Synthase Kinase 3 beta , Gonadal Steroid Hormones , Mice , Proto-Oncogene Proteins c-akt , Receptors, Prostaglandin E, EP2 Subtype/metabolism , Receptors, Prostaglandin E, EP4 Subtype , beta CateninABSTRACT
Advances in next-generation sequencing and other high-throughput technologies have facilitated multiomics research, such as genomics, epigenomics, transcriptomics, proteomics, metabolomics, and phenomics. The resultant emerging multiomics data have brought new challenges as well as opportunities, as seen in the plant and agriculture science domains. We reviewed several bioinformatic and computational methods, models, and platforms, and we have highlighted some of our in-house developed efforts aimed at multiomics data analysis, integration, and management issues faced by the research community. A case study using multiomics datasets generated from our studies of maize nodal root growth under water deficit stress demonstrates the power of these datasets and some other publicly available tools. This analysis also sheds light on the landscape of such applied bioinformatic tools currently available for plant and crop science studies and introduces emerging trends and how they may affect the future.
Subject(s)
Computational Biology , Zea mays , Agriculture , Computational Biology/methods , Genomics/methods , Plants , Water , Zea mays/geneticsABSTRACT
Ambient carbon monoxide (CO) is associated with bronchitis morbidity, but there is no evidence concerning its correlation with hospitalization costs for bronchitis patients. This study aimed to investigate the relationship between short-term ambient CO exposure and hospitalization costs for bronchitis patients in Chongqing, China. Baseline data for 3162 hospitalized bronchitis patients from November 2013 to December 2019 were collected. Multiple linear regression analysis was used to determine the association, delayed and cumulative, between short-term CO exposure and hospitalization costs. Additionally, subgroup analyses were performed by gender, age, season, and comorbidity. Positive association between CO and hospitalization costs for bronchitis patients was observed. The strongest association was observed at lag 015 days, with per 1 mg/m3 increase of CO concentrations corresponded to 5834.40 Chinese Yuan (CNY) (95% CI: 2318.71, 9350.08; P < 0.001) (845.97 US dollars) increment in hospitalization costs. Stratified analysis results showed that the association was more obvious among those males, elderly, with comorbidities, and in warm seasons. More importantly, there was strongest correlation between CO and bronchitis patients with coronary heart disease. In summary, short-term exposure to ambient CO, even lower than Chinese and WHO standards, can be associated with increased hospitalization costs for bronchitis. Controlling CO exposure can be helpful to reduce medical burden associated with bronchitis patients. The results also suggest that when setting air quality standards and formulating preventive measures, susceptible subpopulations ought to be considered.
Subject(s)
Air Pollutants , Air Pollution , Bronchitis , Aged , Air Pollutants/analysis , Air Pollutants/toxicity , Air Pollution/analysis , Bronchitis/epidemiology , Carbon Monoxide/analysis , China/epidemiology , Environmental Exposure/analysis , Hospitalization , Hospitals , Humans , Male , Particulate Matter/analysisABSTRACT
Pancreatic cancer (PaC) is resistant to immune checkpoint therapy, but the underlying mechanisms are largely unknown. In this study, we have established four orthotopic PaC murine models with different PaC cell lines by intra-pancreatic inoculation. Therapeutic examinations demonstrate that only tumors induced with Panc02-H7 cells respond to αPD-1 antibody treatment, leading to significantly reduced tumor growth and increased survival in the recipient mice. Transcriptomic profiling at a single-cell resolution characterizes the molecular activity of different cells within tumors. Comparative analysis and validated experiments demonstrate that αPD-1-sensitive and -resistant tumors differently shape the immune landscape in the tumor microenvironment (TME) and markedly altering effector CD8+ T cells and tumor-associated macrophages (TAMs) in their number, frequency, and gene profile. More exhausted effector CD8+ T cells and increased M2-like TAMs with a reduced capacity of antigen presentation are detected in resistant Panc02-formed tumors versus responsive Panc02-H7-formed tumors. Together, our data highlight the correlation of tumor-induced imbalance of macrophages with the fate of tumor-resident effector CD8+ T cells and PaC response to αPD-1 immunotherapy. TAMs as a critical regulator of tumor immunity and immunotherapy contribute to PaC resistance to immune checkpoint blockade.
ABSTRACT
The accurate annotation of protein localization is crucial in understanding protein function in tandem with a broad range of applications such as pathological analysis and drug design. Since most proteins do not have experimentally-determined localization information, the computational prediction of protein localization has been an active research area for more than two decades. In particular, recent machine-learning advancements have fueled the development of new methods in protein localization prediction. In this review paper, we first categorize the main features and algorithms used for protein localization prediction. Then, we summarize a list of protein localization prediction tools in terms of their coverage, characteristics, and accessibility to help users find suitable tools based on their needs. Next, we evaluate some of these tools on a benchmark dataset. Finally, we provide an outlook on the future exploration of protein localization methods.
ABSTRACT
Prediction of protein localization plays an important role in understanding protein function and mechanisms. In this paper, we propose a general deep learning-based localization prediction framework, MULocDeep, which can predict multiple localizations of a protein at both subcellular and suborganellar levels. We collected a dataset with 44 suborganellar localization annotations in 10 major subcellular compartments-the most comprehensive suborganelle localization dataset to date. We also experimentally generated an independent dataset of mitochondrial proteins in Arabidopsis thaliana cell cultures, Solanum tuberosum tubers, and Vicia faba roots and made this dataset publicly available. Evaluations using the above datasets show that overall, MULocDeep outperforms other major methods at both subcellular and suborganellar levels. Furthermore, MULocDeep assesses each amino acid's contribution to localization, which provides insights into the mechanism of protein sorting and localization motifs. A web server can be accessed at http://mu-loc.org.
ABSTRACT
Microbe associated molecular pattern (MAMPs) triggered immunity (MTI) is a key component of the plant innate immunity response to microbial recognition. However, most of our current knowledge of MTI comes from model plants (i.e., Arabidopsis thaliana) with comparatively less work done using crop plants. In this work, we studied the MAMP triggered oxidative burst (ROS) and the transcriptional response in two Sorghum bicolor genotypes, BTx623 and SC155-14E. SC155-14E is a line that shows high anthracnose resistance and the line BTx623 is susceptible to anthracnose. Our results revealed a clear variation in gene expression and ROS in response to either flagellin (flg22) or chitin elicitation between the two lines. While the transcriptional response to each MAMP and in each line was unique there was a considerable degree of overlap, and we were able to define a core set of genes associated with the sorghum MAMP transcriptional response. The GO term and KEGG pathway enrichment analysis discovered more immunity and pathogen resistance related DEGs in MAMP treated SC155-14E samples than in BTx623 with the same treatment. The results provide a baseline for future studies to investigate innate immunity pathways in sorghum, including efforts to enhance disease resistance.
ABSTRACT
Systemic inflammation in pregnant obese women is associated with 1.5- to 2-fold increase in serum Interleukin-6 (IL-6) and newborns with lower kidney/body weight ratio but the role of IL-6 in increased susceptibility to chronic kidney (CKD) in adult progeny is not known. Since IL-6 crosses the placental barrier, we administered recombinant IL-6 (10 pg/g) to pregnant mice starting at mid-gestation yielded newborns with lower body (p < 0.001) and kidney (p < 0.001) weights. Histomorphometry indicated decreased nephrogenic zone width (p = 0.039) with increased numbers of mature glomeruli (p = 0.002) and pre-tubular aggregates (p = 0.041). Accelerated maturation in IL-6 newborns was suggested by early expression of podocyte-specific protein podocin in glomeruli, increased 5-methyl-cytosine (LC-MS analysis for CpG DNA methylation) and altered expression of certain genes of cell-cycle and apoptosis (RT-qPCR array-analysis). Western blotting showed upregulated pJAK2/pSTAT3. Thus, treating dams with IL-6 as a surrogate provides newborns to study effects of maternal systemic inflammation on future susceptibility to CKD in adulthood.
Subject(s)
Interleukin-6/adverse effects , Prenatal Exposure Delayed Effects/chemically induced , Animals , Animals, Newborn/growth & development , Apoptosis/drug effects , Birth Weight/drug effects , Cell Cycle/drug effects , Female , Kidney/growth & development , Kidney/metabolism , Kidney/pathology , Male , Mice , Mice, Inbred C57BL , Organ Size/drug effects , Pregnancy , Prenatal Exposure Delayed Effects/pathologyABSTRACT
Increased fluid flow shear stress (FFSS) in solitary kidney alters podocyte function in vivo. FFSS-treated cultured podocytes show upregulated AKT-GSK3ß-ß-catenin signaling. The present study was undertaken to confirm (i) the activation of ß-catenin signaling in podocytes in vivo using unilaterally nephrectomized (UNX) TOPGAL mice with the ß-galactosidase reporter gene for ß-catenin activation, (ii) ß-catenin translocation in FFSS-treated mouse podocytes, and (iii) ß-catenin signaling using publicly available data from UNX mice. The UNX of TOPGAL mice resulted in glomerular hypertrophy and increased the mesangial matrix consistent with hemodynamic adaptation. Uninephrectomized TOPGAL mice showed an increased ß-galactosidase expression at 4 weeks but not at 12 weeks, as assessed using immunofluorescence microscopy (p < 0.001 at 4 weeks; p = 0.16 at 12 weeks) and X-gal staining (p = 0.008 at 4 weeks; p = 0.65 at 12 weeks). Immunofluorescence microscopy showed a significant increase in phospho-ß-catenin (Ser552, p = 0.005) at 4 weeks but not at 12 weeks (p = 0.935) following UNX, and the levels of phospho-ß-catenin (Ser675) did not change. In vitro FFSS caused a sustained increase in the nuclear translocation of phospho-ß-catenin (Ser552) but not phospho-ß-catenin (Ser675) in podocytes. The bioinformatic analysis of the GEO dataset, #GSE53996, also identified ß-catenin as a key upstream regulator. We conclude that transcription factor ß-catenin mediates FFSS-induced podocyte (glomerular) injury in solitary kidney.
Subject(s)
Glomerular Filtration Rate , Mechanotransduction, Cellular , Podocytes/metabolism , Solitary Kidney/metabolism , beta Catenin/metabolism , Animals , Cell Line , Databases, Genetic , Disease Models, Animal , Genes, fos , Lac Operon , Lymphoid Enhancer-Binding Factor 1/genetics , Mice, Transgenic , Podocytes/pathology , Promoter Regions, Genetic , Solitary Kidney/genetics , Solitary Kidney/pathology , Solitary Kidney/physiopathology , Stress, Mechanical , Transcription Factor 3/genetics , beta Catenin/geneticsABSTRACT
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 Alzheimer's 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 powerful framework that can be applied to general scRNA-Seq analyses.