Biology subjects

Zeng, W.

Publications and source records attributed to Zeng, W..

At least 55 records · Page 3Linked to original sources

bioRxiv2024

Virological characteristics of SARS-CoV-2 Omicron BA.5.2.48

With the prevalence of sequentially-emerged sublineages including BA.1, BA.2 and BA.5, SARS-CoV-2 Omicron infection has transformed into a regional epidemic disease. As a sublineage of BA.5, the BA.5.2.48 outbreak and evolved into multi-subvariants in China without clearly established virological characteristics, especially the pathogenicity. Though reduced airborne transmission and pathogenicity of former Omicron sublineages have been revealed in animal models, the virological characteristics of BA.5.2.48 was unidentified. Here, we evaluated the in vitro and in vivo virological characteristics of two isolates of the prevalent BA.5.2.48 subvariant, DY.2 and DY.1.1 (a subvariant of DY.1). DY.2 replicates more efficiently than DY.1.1 in HelahACE2+ cells and Calu-3 cells. The A570S mutation (of DY.1) in a normal BA.5 spike protein (DY.2) leads to a 20% improvement in the hACE2 binding affinity, which is slightly reduced by a further K147E mutation (of DY.1.1). Compared to the normal BA.5 spike, the double-mutated protein demonstrates efficient cleavage and reduced fusogenicity. BA.5.2.48 demonstrated enhanced airborne transmission capacity in hamsters than BA.2. The pathogenicity of BA.5.2.48 is greater than BA.2, as revealed in K18-hACE2 rodents. Under immune selection pressure, DY.1.1 shows stronger fitness than DY.2 in hamster turbinates. Thus the outbreaking prevalent BA.5.2.48 multisubvariants exhibites divergent virological features. ImportanceOmicron continues to circulate and evolves novel sublineages with indistinguishable pathogenicity and transmission. Therefore humanized Omicron-sensitive animal models must be applied to evaluate the virological characteritics and antiviral therapeutics. By using multiple models including the Omicron-lethal H11-K18-hACE2 rodents, BA.5.2.48 revealed higher pathogenicity in the novel H11-K18-hACE2 rodent models than the previously epidemic BA.2, and thus the models are more adapted to Omicron studies. Moreover, the regional outbreaking of BA.5.2.48 promotes the multidirectional evolution of its subvariants, gaining either enhanced pathogenicity or a fitness in upper airways which is associated with higher transmission, highlighting the importance of surveillance and virological studies on regionally endemic sublineages which represents the short-run evolutionary direction of Omicron.

microbiology↗

bioRxiv2024

Characterization of the SARS-CoV-2 BA.5 Variants in H11-K18-hACE2 Hamsters

This study aims to comprehensively characterize the SARS-CoV-2 BA.5 variants using K18 hACE2 transgenic mice and golden hamsters as model organisms. Previous research on SARS-CoV-2 has utilized both mouse and hamster models, leading to conflicting results concerning the viruss lethality. In our study, the finding suggests that H11-K18 hACE2 golden hamsters closely mimic the disease progression observed in human COVID-19 cases caused by BA.5 variants, demonstrating consistent severity and symptoms comparable to severe infections. Additionally, hamsters exhibit heightened respiratory viral replication, accurately reflecting the clinical viral kinetics observed in humans. The study emphasizes the critical importance of selecting an appropriate animal model for SARS-CoV-2 research, while also providing robust support for the hypothesis that BA.5 variants contribute to fatal outcomes in COVID-19 cases. These findings highlight the pivotal role of the golden hamster model in advancing our understanding of the pathogenic mechanisms underlying SARS-CoV-2 variants, as well as in the development of targeted therapeutic strategies. Significance StatementOur research work explores groundbreaking insights that could reshape our understanding of COVID-19 and pave the way for targeted therapies. We use golden hamster models to express the possibility of different animal models could contribute to human diseases. We hope this finding could clarify some conflicts existed, and help further development of medication for COVID.

immunology↗

bioRxiv2023

Cultivation of host-adapted Cryptosporidium parvum and Cryptosporidium hominis using enteroids for cryopreservation of isolates and transcriptomic studies of infection

Cryptosporidium hominis and Cryptosporidium parvum are major causes of severe diarrhea in humans. Comparative studies of them are hampered by the lack of effective cultivation and cryopreservation methods, especially for C. hominis. Here, we described adapted murine enteroids for the cultivation of one C. parvum IId subtype and nonhuman primate-adapted C. hominis Ib, Im, and In subtypes, which allowed the complete development of the pathogens, producing oocysts infectious to mice. Using the system, we developed a novel cryopreservation method for Cryptosporidium isolates. In comparative RNA-seq analyses of C. hominis cultures, the enteroid system generated significantly more transcriptomic responses of both pathogen and host genes than the conventional HCT-8 cell system. In particular, the infection was shown to upregulate PI3K-Akt, Wnt, Ras,TNF, NF-{kappa}B, IL-17, MAPK, and innate immunity signaling pathways and downregulate Wnt and Hippo signaling pathways, host cell metabolism, and parasites in enteroid cultures had significantly higher expression of genes involved in oocyst formation. Therefore, the new culture model provides a valuable tool for comparative studies of the biology of divergent Cryptosporidium species. IMPORTANCE The two dominant species for human cryptosporidiosis, Cryptosporidium hominis and Cryptosporidium parvum, differ significantly in host range and virulence. Up to date, biological studies of Cryptosporidium spp. are almost exclusively done with bovine-adapted IIa subtypes of C. parvum, which is the species with effective laboratory animal models and in vitro cultivation methods. Here, we describe modified procedures for the generation of murine enteroids for successful cultivation of both nonhuman primate-adapted C. hominis subtypes and a C. parvum IId subtype, producing oocysts infective to mice. In addition, we have developed a novel cryopreservation method using the system for long-term storage of Cryptosporidium isolates. RNA-seq analyses of C. hominis cultures indicate that the enteroid culture system generates host and pathogen transcriptomic responses similar to those in natural infection. This new development alleviates a technical bottleneck in cryptosporidiosis research, and provides an example for other difficult-to-culture pathogens of major public health importance.

microbiology↗

bioRxiv2023

EpiGePT: a Pretrained Transformer model for epigenomics

The inherent similarities between natural language and biological sequences have given rise to great interest in adapting the transformer-based large language models (LLMs) underlying recent breakthroughs in natural language processing (references), for applications in genomics. However, current LLMs for genomics suffer from several limitations such as the inability to include chromatin interactions in the training data, and the inability to make prediction in new cellular contexts not represented in the training data. To mitigate these problems, we propose EpiGePT, a transformer-based pretrained language model for predicting context-specific epigenomic signals and chromatin contacts. By taking the context-specific activities of transcription factors (TFs) and 3D genome interactions into consideration, EpiGePT offers wider applicability and deeper biological insights than models trained on DNA sequence only. In a series of experiments, EpiGePT demonstrates superior performance in a diverse set of epigenomic signals prediction tasks when compared to existing methods. In particular, our model enables cross-cell-type prediction of long-range interactions and offers insight on the functional impact of genetic variants under different cellular contexts. These new capabilities will enhance the usefulness of LLM in the study of gene regulatory mechanisms. We provide free online prediction service of EpiGePT through http://health.tsinghua.edu.cn/epigept/.

genomics↗

bioRxiv2023

MR-DNA: Flexible 5mC-Methylation-Site Recognition in DNA Sequences using Token Classification

DNA 5-methylcytosine modification has been widely studied in mammals and plays an important role in epigenetics. Several computational methods exist that attempt to determine the methylation state of a DNA sequence centered at a possible methylation site. Here, we introduce a novel deep-learning framework, MR-DNA, that predicts the methylation state of a single nucleotide located in a gene promoter region. The idea is to adapt the named-entity recognition approach to methylation-site prediction and to incorporate biological rules during model construction. MR-DNA has a stacked model architecture consisting of a pre-trained MuLan-Methyl-DistilBERT language model and a conditional random field algorithm, trained with a self-defined methyl loss function. The resulting fine-tuned model achieves an accuracy of 97.9% on an independent test dataset of samples. An advantage of this formulation of the methylation-site identification task is that it predicts on every nucleotide of a sequence of a given length, unlike previous methods that the predict methylation state of DNA sequences of a short fixed length. For training and testing purposes, we provide a database of DNA sequences containing verified 5mC-methylation sites, calculated from data for eight human cell lines downloaded from the ENCODE database.

bioinformatics↗

bioRxiv2023

The ENCODE4 long-read RNA-seq collection reveals distinct classes of transcript structure diversity

The majority of mammalian genes encode multiple transcript isoforms that result from differential promoter use, changes in exonic splicing, and alternative 3 end choice. Detecting and quantifying transcript isoforms across tissues, cell types, and species has been extremely challenging because transcripts are much longer than the short reads normally used for RNA-seq. By contrast, long-read RNA-seq (LR-RNA-seq) gives the complete structure of most transcripts. We sequenced 264 LR-RNA-seq PacBio libraries totaling over 1 billion circular consensus reads (CCS) for 81 unique human and mouse samples. We detect at least one full-length transcript from 87.7% of annotated human protein coding genes and a total of 200,000 full-length transcripts, 40% of which have novel exon junction chains. To capture and compute on the three sources of transcript structure diversity, we introduce a gene and transcript annotation framework that uses triplets representing the transcript start site, exon junction chain, and transcript end site of each transcript. Using triplets in a simplex representation demonstrates how promoter selection, splice pattern, and 3 processing are deployed across human tissues, with nearly half of multitranscript protein coding genes showing a clear bias toward one of the three diversity mechanisms. Evaluated across samples, the predominantly expressed transcript changes for 74% of protein coding genes. In evolution, the human and mouse transcriptomes are globally similar in types of transcript structure diversity, yet among individual orthologous gene pairs, more than half (57.8%) show substantial differences in mechanism of diversification in matching tissues. This initial large-scale survey of human and mouse long-read transcriptomes provides a foundation for further analyses of alternative transcript usage, and is complemented by short-read and microRNA data on the same samples and by epigenome data elsewhere in the ENCODE4 collection.

genomics↗

bioRxiv2023

RBP-J regulates homeostasis and function of circulating Ly6C lo monocytes

Notch-RBP-J signaling plays an essential role in maintenance of myeloid homeostasis. However, its role in monocyte cell fate decisions is not fully understood. Here we showed that conditional deletion of transcription factor RBP-J in myeloid cells resulted in marked accumulation of blood Ly6Clo monocytes that highly expressed chemokine receptor CCR2. Bone marrow transplantation and parabiosis experiments revealed a cell intrinsic requirement of RBP-J for controlling blood Ly6CloCCR2hi monocytes. RBP-J-deficient Ly6Clo monocytes exhibited enhanced capacity competing with wildtype counterparts in blood circulation. In accordance with alterations of circulating monocytes, RBP-J deficiency led to markedly increased population of lung tissues with Ly6Clo monocytes and CD16.2+ interstitial macrophages. Furthermore, RBP-J deficiency-associated phenotypes could be genetically corrected by further deleting Ccr2 in myeloid cells. These results demonstrate that RBP-J functions as a crucial regulator of blood Ly6Clo monocytes and thus derived lung-resident myeloid populations, at least in part through regulation of CCR2.

immunology↗

bioRxiv2023

Microbiome Metabolome Integration Platform (MMIP): a web-based platform for microbiome and metabolome data integration and feature identification.

A microbial community maintains its ecological dynamics via metabolite crosstalk. Hence knowledge of the metabolome, alongside its populace, would help us understand the functionality of a community and also predict how it will change in atypical conditions. Methods that employ low-cost metagenomic sequencing data can predict the metabolic potential of a community, that is, its ability to produce or utilize specific metabolites. These, in turn, can potentially serve as markers of biochemical pathways that are associated with different communities. We developed MMIP (Microbiome Metabolome Integration Platform), a web-based analytical and predictive tool that can be used to compare the taxonomic content, diversity variation and the metabolic potential between two sets of microbial communities from targeted amplicon sequencing data. MMIP is capable of highlighting statistically significant taxonomic, enzymatic and metabolic attributes as well as learning-based features associated with one group in comparison with another. Further MMIP can predict linkages among species or groups of microbes in the community, specific enzyme profiles, compounds or metabolites associated with such a group of organisms. With MMIP, we aim to provide a user-friendly, online web-server for performing key microbiome-associated analyses of targeted amplicon sequencing data, predicting metabolite signature, and using learning-based linkage analysis, without the need for initial metabolomic analysis, and thereby helping in hypothesis generation.

bioinformatics↗

bioRxiv2023

Homocysteine accelerates hepatocyte autophagy by upregulation of TFEB via DNMT3b-mediated DNA hypomethylation

Autophagy plays a critical role in the physiology and pathophysiology of hepatocytes. High levels of homocysteine (Hcy) promote autophagy in hepatocytes, but the underlying mechanism is still unknown. Here, we investigated the relation between Hcy increased autophagy levels and the expression of nuclear transcription factor EB (TFEB). We demonstrate that Hcy increased autophagy levels is mediated by upregulation of TFEB. Silencing TFEB decreases the autophagy-related protein LC3BII/I and increases p62 expression levels in hepatocytes after exposure to Hcy. Moreover, the effect of Hcy on the expression of TFEB is regulated by hypomethylation of TFEB promoter catalyzed by DNA methyltransferase 3b (DNMT3b). In summary, this study shows that Hcy can activate autophagy by inhibiting DNMT3b-mediated DNA methylation and upregulating TFEB expression. These findings provide another new mechanism for Hcy-induced autophagy in hepatocytes.

cell biology↗

bioRxiv2023

MuLan-Methyl - Multiple Transformer-based Language Models for Accurate DNA Methylation Prediction

Transformer-based language models are successfully used to address massive text-related tasks. DNA methylation is an important epigenetic mechanism and its analysis provides valuable insights into gene regulation and biomarker identification. Several deep learning-based methods have been proposed to identify DNA methylation and each seeks to strike a balance between computational effort and accuracy. Here, we introduce MuLan-Methyl, a deep-learning framework for predicting DNA methylation sites, which is based on five popular transformer-based language models. The framework identifies methylation sites for three different types of DNA methylation, namely N6-adenine, N4-cytosine, and 5-hydroxymethylcytosine. Each of the employed language models is adapted to the task using the "pre-train and fine-tune" paradigm. Pre-training is performed on a custom corpus of DNA fragments and taxonomy lineages using self-supervised learning. Fine-tuning aims at predicting the DNA-methylation status of each type. The five models are used to collectively predict the DNA methylation status. We report excellent performance of MuLan-Methyl on a benchmark dataset. Moreover, we argue that the model captures characteristic differences between different species that are relevant for methylation. This work demonstrates that language models can be successfully adapted to applications in biological sequence analysis and that joint utilization of different language models improves model performance. Mulan-Methyl is open source and we provide a web server that implements the approach. Key pointsO_LIMuLan-Methyl aims at identifying three types of DNA-methylation sites. C_LIO_LIIt uses an ensemble of five transformer-based language models, which were pre-trained and fine-tuned on a custom corpus. C_LIO_LIThe self-attention mechanism of transformers give rise to importance scores, which can be used to extract motifs. C_LIO_LIThe method performs favorably in comparison to existing methods. C_LIO_LIThe implementation can be applied to chromosomal sequences to predict methylation sites. C_LI

bioinformatics↗

bioRxiv2022

Megan Server: facilitating interactive access to metagenomic data on a server

MotivationMetagenomic projects of large sequencing datasets (totaling hundreds of gigabytes of data). Thus, computational preprocessing and analysis are usually performed on a server rather than on a personal computer. The results of such analyses are then usually explored interactively. One approach is to use MEGAN, an interactive program that allows analysis and comparison of metagenomic datasets. Previous releases of MEGAN have required that the user first download the computed data from the server, an increasingly time-consuming process. Here we present Megan Server, a stand-alone program that serves MEGAN files to the web, using a RESTful API, facilitating interactive analysis without downloading the complete data. We describe a number of different application scenarios. AvailabilityMegan Server is provided as a standalone program tools/megan-server in the Megan6 software suite, available at https://software-ab.cs.uni-tuebingen.de/download/megan6.

bioinformatics↗

bioRxiv2022

Complexes Formed by Mixing Tumor-derived Extracellular Vesicles with Polymeric Surfactants for Personalized Therapeutic Vaccine

The personalized therapeutic vaccine is an ideal weapon to eliminate tumors. However, the core steps of manufacturing personalized cancer vaccines are identifying tumor-specific antigens (TSAs, also called neoantigens) and HLA epitope prediction, which is time-consuming and labor-intense. Tumor-derived extracellular vesicles (TEVs) are alternative sources of neoantigens. However, the immunosuppressive nature of TEVs limits their application in such immunotherapy. In this study, we present a new strategy to maintain neoantigens in TEVs and diminish the immunosuppression by deconstructing the structure of TEVs with polymeric surfactant polyethylene glycol-phosphatidylethanolamine (PEG-PE). Together with adjuvant MPLA, the newly formed micelle-like complexes compose a therapeutic vaccine (MLC-V). Results show that MLC-V is capable of eliciting neoantigen-specific T-cell responses, restoring TEV-induced immunosuppression, and preventing lung metastasis of murine melanoma. MLC-V also exhibits outstanding anti-tumor efficacy in multiple tumor models. MLC-V can be used as a personalized therapeutic vaccine in a mimetic pre-clinical MC38 model and the anti-tumor effect of MLC-V was synergistically enhanced by PD-1 mAb. Taken together, the present study demonstrates a time-saving, low-cost, and simplified strategy to produce personalized therapeutic vaccines based on MLC-V platform technology.

immunology↗

bioRxiv2022

Alternative mRNA polyadenylation bridges mitochondrion-to-nucleus communication in Arabidopsis

Mitochondria produce signals besides energy and metabolites that influence plant growth and fitness. However, how mitochondrial signals are relayed to other cellular compartments is largely unknown. By applying poly(A)-site RNA-sequencing (PAS-seq) to wildtype Arabidopsis seedlings and a mutant in the histone demethylase JMJ30 treated with the mitochondrial electron transfer chain inhibitor antimycin A (AA), we identified a previously undefined mitochondrion-to-nucleus communication pathway by which mitochondrial functional state regulates co-transcriptionally alternative polyadenylation (APA) of nuclear mRNA. We observed a global shortening of 3' untranslated regions (UTRs) as a molecular signature of AA-activated mitochondrial retrograde response (MRR), which contributed in part to translational regulation of auxin response and cell wall biogenesis. JMJ30 regulated AA-induced 3' UTR shortening, resulting in more transcripts with shortened 3' UTRs upon AA treatment in a JMJ30 gain-of-function mutant and overexpression lines. We also report on the JMJ30-interacting protein CPSF30, a cleavage and polyadenylation specificity factor that recruits JMJ30 to modulate H3K27me3 status at its target loci. Our study illustrates how epigenetic modification and APA coordinate mitochondrion-to-nucleus communication to allow cells to rapidly respond to changes in mitochondrial functional state and shape plant growth and fitness. One-sentence summaryEpigenetic modification and APA coordinate mitochondrion-to-nucleus communication to allow cells to rapidly respond to changes in mitochondrial functional state and shape plant growth and fitness.

molecular biology↗

bioRxiv2022

Deep generative modeling and clustering of single cell Hi-C data

Deciphering 3D genome conformation is important for understanding gene regulation and cellular function at a spatial level. The recent advances of single cell Hi-C technologies have enabled the profiling of the 3D architecture of DNA within individual cell, which allows us to study the cell-to-cell variability of 3D chromatin organization. Computational approaches are in urgent need to comprehensively analyze the sparse and heterogeneous single cell Hi-C data. Here, we proposed scDEC-Hi-C, a new framework for single cell Hi-C analysis with deep generative neural networks. scDEC-Hi-C outperforms existing methods in terms of single cell Hi-C data clustering and imputation. Moreover, the generative power of scDEC-Hi-C could help unveil the heterogeneity of chromatin architecture across different cell types. We expect that scDEC-Hi-C could shed light on deepening our understanding of the complex mechanism underlying the formation of chromatin contacts. scDEC-Hi-C is freely available at https://github.com/kimmo1019/scDEC-Hi-C. Key pointsO_LIscDEC-Hi-C provides an end-to-end framework based on autoencoder and deep generative model to comprehensively analyze single cell Hi-C data, including low-dimensional embedding and clustering. C_LIO_LIThrough a series of experiments including single cell Hi-C data clustering and structural difference identification, scDEC-Hi-C demonstrates suprioir performance over existing methods. C_LIO_LIIn the downstream analysis of chromatin loops from single cell Hi-C data, scDEC-Hi-C is capable of significantly enhancing the ability for identifying single cell chromatin loops by data imputation. C_LI

bioinformatics↗

bioRxiv2022

CARD8 negatively regulates NLRP1 inflammasome activation level by interaction with NLRP1

NLRP1 inflammasome functions as canonical cytosolic sensor in response to intracellular infections and is implicated in auto-inflammatory diseases. But the regulation and signal transduction mechanisms of NLRP1 are incompletely understood. Here, we show that the T60 variant of CARD8, but not the canonical T48 isoform, negatively regulates NLRP1 inflammasome activation by directly interacting with the receptor molecule NLRP1 and inhibiting inflammasome assembly. Furthermore, our results suggest that the different ASC preference in three types of inflammasomes, namely ASC-indispensable NLRP1 inflammasome, ASC-dispensable mNLRP1b inflammasome and ASC-independent CARD8 inflammasome, is mainly caused by the CARD domain, not the UPA subdomain. Based on the systematic site-directed mutagenesis and structural analysis, we find that the signal transduction of NLRP1 inflammasome relies on multiple interaction surfaces on its death domain superfamily member CARD domain. Finally, our results partly explain the mechanism of the NLRP1 mutation-derived auto-inflammatory diseases caused by the overactivation of the NLRP1 inflammasome. In conclusion, our study not only reveals how CARD8 downregulates NLRP1 inflammasome activation, but also provides insights into the mechanisms of CARD-containing inflammasome assembly.

immunology↗

bioRxiv2022

Specific binding of Hsp27 and phosphorylated Tau mitigates abnormal Tau aggregation-induced pathology

Amyloid aggregation of phosphorylated Tau (pTau) into neurofibrillary tangles is closely associated with Alzheimers disease (AD). Several molecular chaperones have been reported to bind Tau and impede its pathological aggregation. Recent findings of elevated levels of Hsp27 in the brains of patients with AD suggested its important role in pTau pathology. However, the molecular mechanism of Hsp27 in pTau aggregation remains poorly understood. Here, we show that Hsp27 partially co-localizes with pTau tangles in the brains of patients with AD. Notably, phosphorylation of Tau by microtubule affinity regulating kinase 2 (MARK2), dramatically enhances the binding affinity of Hsp27 to Tau. Moreover, Hsp27 efficiently prevents pTau fibrillation in vitro and mitigates neuropathology of pTau aggregation in a Drosophila tauopathy model. Further mechanistic study reveals that Hsp27 employs its N-terminal domain to directly interact with multiple phosphorylation sites of pTau for specific binding. Our work provides the structural basis for the specific recognition of Hsp27 to pathogenic pTau, and highlights the important role of Hsp27 in preventing abnormal aggregation and pathology of pTau in AD.

biochemistry↗

bioRxiv2022

WDR47 facilitates cell-specific ciliogenesis by modulating intraflagellar transport

Cilia are conserved organelles found in many cell types in eukaryotes, and their dysfunction causes defects in environmental sensing and signaling transduction; such defects are termed ciliopathies. Distinct cilia have cell-specific morphologies and exert distinct functions. However, the underlying mechanisms of cell-specific ciliogenesis and regulation are unclear. Here we identified a WD40-repeat (WDR) protein, WDR47/NMTN-1, and show that it is specifically required for ciliogenesis of AWB chemosensory neurons in C. elegans. WDR47/NMTN-1 is expressed in the AWB chemosensory neuron pair, and is localized at the basal body (BB) of the AWB cilia. Knockout of wdr47/nmtn-1 causes abnormal AWB neuron cilia morphology, structural integrity, and induces aberrant AWB-mediated aversive behaviors. We further demonstrate that wdr47/nmtn-1 deletion affects movement of intraflagellar transport (IFT) particles and their cargo delivery in AWB neurons. Our results indicate that WDR47/NMTN-1 is essential for AWB neuron ciliary morphology and function, which reveal a novel mechanism for cell-specific ciliogenesis. Since WDR47/NMTN-1 is conserved in mammals, our findings may help understand the process of cell-specific ciliogenesis and provide insights for treating ciliopathies.

developmental biology↗

bioRxiv2022

DeepToA: An Ensemble Deep-Learning Approach to Predicting the Theater of Activity of a Microbiome

MotivationMetagenomics is the study of microbiomes using DNA sequencing. A microbiome consists of an assemblage of microbes that is associated with a "theater of activity" (ToA). To what degree does the taxonomic and functional content of the former depend on the (details of the) latter? More technically, given a taxonomic and/or functional profile estimated from metagenomic sequencing data, how to predict the associated ToA? Here we present a deep learning approach to this question. We use both taxonomic and functional profiles as input. We apply node2vec to embed hierarchical taxonomic profiles into numerical vectors. We then perform dimension reduction using clustering, to address the sparseness of the taxonomic data and thus make it more amenable to deep learning algorithms. Functional features are combined with textual descriptions of protein families or domains. We present an ensemble deep-learning framework DeepToA for predicting the theater of activity" of microbial community, based on taxonomic and functional profiles. We use SHAP (SHapley Additive exPlanations) values to determine which taxonomic and functional features are important for the prediction. ResultsBased on 7,560 metagenomic profiles downloaded from MGnify, classified into ten different theaters of activity, we demonstrate that DeepToA has an accuracy of 98.61%. We show that adding textual information to functional features increases the accuracy. AvailabilityOur approach is available at http://ab.inf.uni-tuebingen.de/software/deeptoa. Contactdaniel.huson@uni-tuebingen.de Supplementary informationSupplementary data are available at Bioinformatics online.

bioinformatics↗