Biology subjects

Xu, H.

Publications and source records attributed to Xu, H..

At least 253 records · Page 14Linked to original sources

bioRxiv2023

Adaptor Protein-3 Produces Synaptic Vesicles that Release Phasic Dopamine

The burst firing of midbrain dopamine neurons releases a phasic dopamine signal that mediates reinforcement learning. At many synapses, however, high firing rates deplete synaptic vesicles (SVs), resulting in synaptic depression that limits release. What accounts for the increased release of dopamine by stimulation at high frequency? We find that adaptor protein-3 (AP-3) and its coat protein VPS41 promote axonal dopamine release by targeting vesicular monoamine transporter VMAT2 to the axon rather than dendrites. AP-3 and VPS41 also produce SVs that respond preferentially to high frequency stimulation, independent of their role in axonal polarity. In addition, conditional inactivation of VPS41 in dopamine neurons impairs reinforcement learning, and this involves a defect in the frequency dependence of release rather than the amount of dopamine released. Thus, AP-3 and VPS41 promote the axonal polarity of dopamine release but enable learning by producing a novel population of SVs tuned specifically to high firing frequency that confers the phasic release of dopamine. Significance statementReinforcement learning requires the phasic dopamine produced by burst firing but synaptic vesicle depletion limits the ability to convey information at high firing rates. We now find that AP-3 has two independent roles in dopamine release. First, AP-3 confers the axonal polarity of dopamine release by targeting vesicular monoamine transporter 2 (VMAT2) to the axon. Second, AP-3 acting locally at the nerve terminal produces synaptic vesicles that respond specifically to high frequency stimulation. Consistent with this, loss of AP-3 impairs reinforcement learning and this reflects the defect in release at high frequency, not the reduction in axonal dopamine.

neuroscience↗

bioRxiv2023

BAC Transgenic Expression of Human TREM2-R47H Remodels Amyloid Plaques but Unable to Reprogram Plaque-associated Microglial Reactivity in 5xFAD Mice

BackgroundGenetic study of late-onset Alzheimers disease (AD) reveals that a rare Arginine-to-Histamine mutation at amino acid residue 47 (R47H) in Triggering Receptor Expressed on Myeloid Cells 2 (TREM2) results in increased disease risk. TREM2 plays critical roles in regulating microglial response to amyloid plaques in AD, leading to their clustering and activation surrounding the plaques. We previously showed that increasing human TREM2 gene dosage exerts neuroprotective effects against AD-related deficits in amyloid depositing mouse models of AD. However, the in vivo effects of the R47H mutation on human TREM2-mediated microglial reprogramming and neuroprotection remains poorly understood. MethodHere we created a BAC transgenic mouse model expressing human TREM2 with the R47H mutation in its cognate genomic context (BAC-TREM2-R47H). Importantly, the BAC used in this study was engineered to delete critical exons of other TREM-like genes on the BAC to prevent confounding effects of overexpressing multiple TREM-like genes. We crossed BAC-TREM2- R47H mice with 5xFAD [1], an amyloid depositing mouse model of AD, to evaluate amyloid pathologies and microglial phenotypes, transcriptomics and in situ expression of key TREM2-dosage dependent genes. We also compared the key findings in 5xFAD/BAC-TREM2-R47H to those observed in 5xFAD/BAC-TREM2 mice. ResultBoth BAC-TREM2 and BAC-TREM2-R47H showed proper expression of three splicing isoforms of TREM2 that are normally found in human. In 5xFAD background, elevated TREM2-R47H gene dosages significantly reduced the plaque burden, especially the filamentous type. The results were consistent with enhanced phagocytosis and altered NLRP3 inflammasome activation in BAC- TREM2-R47H microglia in vitro. However, unlike TREM2 overexpression, elevated TREM2- R47H in 5xFAD failed to ameliorate cognitive and transcriptomic deficits. In situ analysis of key TREM2-dosage dependent genes and microglial morphology uncovered that TREM2-R47H showed a loss-of-function phenotype in reprogramming of plaque-associated microglial reactivity and gene expression in 5xFAD. ConclusionOur study demonstrated that the AD-risk variant has a previously unknown, mixture of partial and full loss of TREM2 functions in modulating microglial response in AD mouse brains. Together, our new BAC-TREM2-R47H model and prior BAC-TREM2 mice are invaluable resource to facilitate the therapeutic discovery that target human TREM2 and its R47H variant to ameliorate AD and other neurodegenerative disorders.

neuroscience↗

bioRxiv2023

SCO-spondin knockout mice exhibit small brain ventricles and mild spine deformation

Reissners fiber (RF) is an extracellular polymer comprising the large monomeric protein SCO-spondin (SSPO) secreted by the subcommissural organ (SCO) that extends through cerebrospinal fluid (CSF)-filled ventricles into the central canal of the spinal cord. In zebrafish, RF and CSF-contacting neurons (CSF-cNs) form an axial sensory system that detects spinal curvature, instructs morphogenesis of the body axis, and enables proper alignment of the spine. In mammalian models, RF has been implicated in CSF circulation. However, challenges in manipulating Sspo, an exceptionally large gene of 15,719 nucleotides, with traditional approaches has limited progress. Here, we generated a Sspo knockout mouse model using CRISPR/Cas9-mediated genome-editing. Sspo knockout mice lacked RF-positive material in the SCO and fibrillar condensates in the brain ventricles. Remarkably, Sspo knockout brain ventricle sizes were reduced compared to littermate controls. Minor defects in thoracic spine curvature were detected in Sspo knockouts, which did not alter basic motor behaviors tested. Altogether, our work in mouse demonstrates that SSPO and RF regulate ventricle size during development but only moderately impact spine geometry.

neuroscience↗

bioRxiv2023

Neurotrophin-3 from the dentate gyrus supports postsynaptic sites of mossy fiber-CA3 synapses and contextual memory

At the center of the hippocampal tri-synaptic loop are synapses formed between mossy fiber (MF) terminals from granule cells in the dentate gyrus (DG) and proximal dendrites of CA3 pyramidal neurons. However, the molecular mechanism regulating the development and function of these synapses is poorly understood. In this study, we showed that neurotrophin-3 (NT3) was expressed in nearly all mature granule cells but not CA3 cells. We selectively deleted the NT3-encoding Ntf3 gene in the DG during the 1st two postnatal weeks to generate a Ntf3 conditional knockout (Ntf3-cKO). Ntf3-cKO mice had normal hippocampal cytoarchitecture but displayed elevated anxiety level and impairments in contextual memory, spatial reference memory and nest building. As MF-CA3 synapses are essential for encoding of contextual memory, we examined synaptic transmission at these synapses using ex vivo electrophysiological recordings. We found that Ntf3-cKO mice showed impaired basal synaptic transmission due to deficits in excitatory postsynaptic currents mediated by AMPA receptors but normal presynaptic function and intrinsic excitability of CA3 pyramidal neurons. Consistent with this selective postsynaptic deficit, Ntf3-cKO mice had fewer and smaller thorny excrescences on proximal apical dendrites of CA3 neurons and lower GluR1 levels in the stratum lucidum area where MF-CA3 synapses reside but normal MF terminals, compared with control mice. Thus, our study indicates that NT3 expressed in the dentate gyrus is crucial for the postsynaptic structure and function of MF-CA3 synapses and hippocampal-dependent memory.

neuroscience↗

bioRxiv2023

Genetic dissection of BDNF and TrkB expression in glial cells

The brain-derived neurotrophic factor (BDNF) and its high-affinity receptor tropomyosin-related kinase receptor B (TrkB) are widely expressed in the central nervous system. It is well documented that neurons express BDNF and full-length TrkB (TrkB.FL), and a lower level of truncated TrkB (TrkB.T). With conflicting results, glial cells also have been reported to express BDNF and TrkB. In the current study, we employed a more sensitive and reliable genetic method to characterize the expression of BDNF and TrkB in glial cells in the mouse brain. We utilized three Cre mouse strains in which Cre recombinase is expressed in the same cells as BDNF, TrkB.FL, or all TrkB isoforms, and crossed them to Cre-dependent EGFP reporter mice to label BDNF- or TrkB-expressing cells. We performed immunohistochemistry with glial cell markers to examine the expression of BDNF and TrkB in microglia, astrocytes, and oligodendrocytes. Surprisingly, we found no BDNF- or TrkB-expressing microglia in the brain and spinal cord. Consistent with previous studies, most astrocytes only express TrkB.T in the adult brain. Moreover, there are a small number of astrocytes and oligodendrocytes that express BDNF, the function of which is to be determined. We also found that oligodendrocyte precursor cells, but not mature oligodendrocytes, express both TrkB.FL and TrkB.T in the adult brain. These results not only clarify the expression of BDNF and TrkB in glial cells, but also open opportunities to investigate previously unidentified roles of BDNF and TrkB in glial cells.

neuroscience↗

bioRxiv2023

Improving the Circularization Efficiency, Stability and Translatability of Circular RNA by circDesign

Synthetic circular RNA (circRNA) has emerged as a promising platform for vaccine and therapeutic development, featuring its uniqueness in a closed-loop structure, cap-independent translation mechanism, and prolonged expression. However, the rational design of a circRNA sequence to jointly improve its stability and protein coding potential remains challenging. In this study, we present circDesign, an efficient algorithm to achieve the optimal design of circRNA by ensuring optimized folding of each segment, which leads to enhanced circularization efficiency, stability, and translatability. Using rabies virus glycoprotein (RABV-G) as the model antigen, we demonstrated that circDesign-generated circRNAs exhibited higher stability and protein translation efficiency in vitro and in vivo compared to other codon adaptation index (CAI)-optimized sequences, thus leading to enhanced in vivo immunogenicity. Ribosome and polysome profiling further revealed that an intact internal ribosome entry site (IRES) structure is critical for efficient translation. By intentionally disrupting the IRES motifs, we observed that the resulting circRNA sequences had lower translatability compared to circDesign-generated sequences. Taken together, our circular RNA design algorithm provides a general strategy to leverage the capability of circRNA as next-generation vaccines or therapeutics.

bioengineering↗

bioRxiv2023

The molecular mechanism and evolutionary divergence of caspase 3/7-regulated gasdermin E activation

Caspase (CASP) is a family of proteases involved in cleavage and activation of gasdermin, the executor of pyroptosis. In human, CASP3 and CASP7 recognize the same consensus motif DxxD, which is present in gasdermin E (GSDME). However, human GSDME is cleaved by CASP3 but not by CASP7. The underlying mechanism of this observation is unclear. In this study, we identified a pyroptotic pufferfish GSDME that was cleaved by both pufferfish CASP3/7 and human CASP3/7. Domain swapping between pufferfish and human CASP and GSDME showed that the GSDME C-terminus and the CASP7 p10 subunit determined the cleavability of GSDME by CASP7. p10 contains a key residue that governs CASP7 substrate discrimination. This key residue is highly conserved in vertebrate CASP3 and in most vertebrate (except mammalian) CASP7. In mammals, the key residue is conserved in non-primate (e.g., mouse) but not in primate. However, mouse CASP7 cleaved human GSDME but not mouse GSDME. These findings revealed the molecular mechanism of CASP7 substrate discrimination and the divergence of CASP3/7-mediated GSDME activation in vertebrate. These results also suggested that mutation-mediated functional alteration of CASP probably enabled the divergence and specialization of different CASP members in the regulation of complex cellular activities in mammals.

evolutionary biology↗

bioRxiv2023

Acyl-CoA thioesterase-2 facilitates beta-oxidation in glycolytic skeletal muscle in a lipid supply dependent manner

Acyl-Coenzyme A (acyl-CoA) thioesters are compartmentalized intermediates that participate in in multiple metabolic reactions within the mitochondrial matrix. The limited availability of free CoA (CoASH) in the matrix raises the question of how the local acyl-CoA concentration is regulated to prevent trapping of CoASH from overload of any specific substrate. Acyl-CoA thioesterase-2 (ACOT2) hydrolyzes long-chain acyl-CoAs to their constituent fatty acids and CoASH, and is the only mitochondrial matrix ACOT refractory to inhibition by CoASH. Thus, we reasoned that ACOT2 may constitutively regulate matrix acyl-CoA levels. Acot2 deletion in murine skeletal muscle (SM) resulted in acyl-CoA build-up when lipid supply and energy demands were modest. When energy demand and pyruvate availability were elevated, lack of ACOT2 activity promoted glucose oxidation. This preference for glucose over fatty acid oxidation was recapitulated in C2C12 myotubes with acute depletion of Acot2, and overt inhibition of {beta}-oxidation was demonstrated in isolated mitochondria from Acot2-depleted glycolytic SM. In mice fed a high fat diet, ACOT2 enabled the accretion of acyl-CoAs and ceramide derivatives in glycolytic SM, and this was associated with worse glucose homeostasis compared to when ACOT2 was absent. These observations suggest that ACOT2 supports CoASH availability to facilitate {beta}-oxidation in glycolytic SM when lipid supply is modest. However, when lipid supply is high, ACOT2 enables acyl-CoA and lipid accumulation, CoASH sequestration, and poor glucose homeostasis. Thus, ACOT2 regulates matrix acyl-CoA concentration in glycolytic muscle, and its impact depends on lipid supply.

biochemistry↗

bioRxiv2023

GIMLET: A Unified Graph-Text Model for Instruction-Based Molecule Zero-Shot Learning

Molecule property prediction has gained significant attention in recent years. The main bottleneck is the label insufficiency caused by expensive lab experiments. In order to alleviate this issue and to better leverage textual knowledge for tasks, this study investigates the feasibility of employing natural language instructions to accomplish molecule-related tasks in a zero-shot setting. We discover that existing molecule-text models perform poorly in this setting due to inadequate treatment of instructions and limited capacity for graphs. To overcome these issues, we propose GIMLET, which unifies language models for both graph and text data. By adopting generalized position embedding, our model is extended to encode both graph structures and instruction text without additional graph encoding modules. GIMLET also decouples encoding of the graph from tasks instructions in the attention mechanism, enhancing the generalization of graph features across novel tasks. We construct a dataset consisting of more than two thousand molecule tasks with corresponding instructions derived from task descriptions. We pretrain GIMLET on the molecule tasks along with instructions, enabling the model to transfer effectively to a broad range of tasks. Experimental results demonstrate that GIMLET significantly outperforms molecule-text baselines in instruction-based zero-shot learning, even achieving closed results to supervised GNN models on tasks such as toxcast and muv.1

bioinformatics↗

bioRxiv2023

Identification of bioactive constituents for colitis from traditional Chinese medicine prescription via deep neural network

Colitis is a commonly encountered inflammatory disease in colon tissue, which can be triggered by various causes. Although traditional Chinese medicine (TCM) has been utilized for the treatment of colitis, it is still a great challenge to identify the major bioactive constituents and their modes of action among thousands of ingredients in TCM prescriptions. Inspired by the success of artificial intelligence and deep learning methods, we proposed a deep neural network (DNN) for TCM prescription recommendation. We constructed a graph-based DNN with 9,845 nodes and 161,950 edges, which integrated microscopic information including bioactive molecules, protein targets, and extracted features of prescriptions through feature embedding. A novel and efficient data augmentation strategy was implemented to expand the sample size based on 378 collected TCM prescriptions. Network pharmacology study revealed that the 10 most frequent ingredients in generated prescriptions were associated with multiple inflammatory signaling pathways. To verify the bioactive constituents in the generated prescriptions, 5 selected constituents were administrated to BALB/c mice with colitis. Suppressive effects of disease progression and pro-inflammatory factors comparable to sulfasalazin were observed with these compounds, revealing the effectiveness of our artificial intelligence strategy on idetification of bioactive constituents from TCM prescriptions.

bioinformatics↗

bioRxiv2023

Identification of unique cytokine expressions in tumor immune microenvironment for various cancer types using single-cell RNA sequencing data via machine learning

Cytokines are small protein molecules that exhibit potent immunoregulatory properties, which are known as the essential components of the tumor immune microenvironment (TIME). While some cytokines are known to be universally upregulated in TIME, the unique cytokine expression patterns have not been fully resolved in specific types of cancers. To address this challenge, we develop a TIME single-cell RNA sequencing (scRNA-seq) dataset, which is designed to study cytokine expression patterns for precise cancer classification. The dataset, including 39 cancers, is constructed by integrating 695 tumor scRNA-seq samples from multiple public repositories. After screening and processing, the dataset retains only the expression data of immune cells. With a machine learning classification model, unique cytokine expression patterns are identified for various cancer categories and pioneering applied to cancer classification with an accuracy rate of 78.01%. Our method will not only boost the understanding of cancer-type-specific immune modulations in TIME but also serve as a crucial reference for future diagnostic and therapeutic research in cancer immunity.

bioinformatics↗

bioRxiv2023

Discovery of new deaminase functions by structure-based protein clustering

The elucidation of protein function and its exploitation in bioengineering have greatly contributed to the development of the life sciences. Existing protein mining efforts generally rely on amino acid sequences rather than protein structures due to technical difficulties in structural elucidation. We describe here for the use of AlphaFold2 to predict and subsequently cluster an entire protein family based on predicted structure similarities. We selected the deaminase family of proteins to analyze and through this approach identified many previously unknown deaminase properties. We applied these new deaminases to the development of new cytosine base editors with distinct features. Although we found many new double-stranded DNA deaminases from the DddA-like protein clade, we were surprised to find that most of the proteins in this family were not actually double-stranded DNA cytidine deaminases. From this protein clade, we engineered the smallest single-strand specific cytidine deaminase, which facilitates the first efficient cytosine base editor to be packaged into a single AAV vector. Importantly, we also profiled a deaminase from this clade that edits robustly in soybean plants, which previously suffered from poor editing by cytosine base editors. These newly discovered deaminases based on AI-assisted structural predictions greatly expand the utility of base editors for therapeutic and agricultural applications.

molecular biology↗

bioRxiv2023

Integrated analysis of spatial multi-omics with SpatialGlue

Integration of multiple data modalities in a spatially informed manner remains an unmet need for exploiting spatial multi-omics data. We introduce SpatialGlue, a graph neural network with dual-attention mechanism, to learn each modalitys significance at cross-omics and intra-omics integration. We demonstrate that SpatialGlue can accurately aggregate cell types into spatial domains at a higher resolution on different tissue types and technology platforms, as well as gain insights into cross-modality spatial correlations.

bioinformatics↗

bioRxiv2023

Site effects depth denoising and signal enhancement using dual-projection based ICA model

Combining magnetic resonance imaging (MRI) data from multi-site studies is a popular approach for constructing larger datasets to greatly enhance the reliability and reproducibility of neuroscience research. However, the scanner/site variability is a significant confound that complicates the interpretation of the results, so effective and complete removal of the scanner/site variability is necessary to realize the full advantages of pooling multi-site datasets. Independent component analysis (ICA) and general linear model (GLM) based denoising methods are the two primary methods used to denoise scanner/site-related effects. Unfortunately, there are challenges with both ICA-based and GLM-based denoising methods to remove site effects completely when the signals of interest and scanner/site-related noises are correlated, which may occur in neuroscience studies. In this study, we propose an effective and powerful denoising strategy that implements dual-projection (DP) theory based on ICA to remove the scanner/site-related effects more completely. This method can separate the signal effects correlated with noise variables from the identified noise effects for removal without losing signals of interest. Both simulations and vivo structural MRI datasets, including a dataset from Autism Brain Imaging Data Exchange II and a traveling subject dataset from the Strategic Research Program for Brain Sciences, were used to test the proposed GLM- and ICA-based denoising methods and our DP-based ICA denoising method. Results show that DP-based ICA

neuroscience↗

bioRxiv2023

Deep flanking sequence engineering for efficient promoter design

Human experts are good at summarizing explicit strong patterns from small samples, while deep learning models can learn implicit weak patterns from big data. Biologists have traditionally described the sequence patterns of promoters via transcription factor binding sites (TFBSs), while the flanking sequences among TFBSs, which can also significantly influence promoter activity and function, remain largely uncharacterized. Thus, current synthetic promoters are mainly designed by the manipulation of TFBSs, while the flanking sequence is often chosen arbitrarily or by previous experience due to a lack of well-summarized optimization criteria. Here, we introduced an AI-aided promoter design framework, DeepSEED, that employs expert knowledge and deep learning methods to efficiently design synthetic promoters that have various desirable functions. DeepSEED incorporates the user-defined cis-regulatory sequences as seeds and generates flanking sequences that match the seeds. We showed that DeepSEED can automatically capture k-mer frequencies and DNA shape features from active promoters in the training set and efficiently optimize the flanking sequences to better match desired properties in synthetic promoters. We validated the effectiveness of this framework for diverse synthetic promoter design tasks in both prokaryotic and eukaryotic cells. DeepSEED successfully designed E. coli constitutive, isopropyl-beta-D-1-thiogalactopyranoside (IPTG)-inducible, and mammalian cell doxycycline (Dox)-inducible promoters with significant performance improvements, suggesting that DeepSEED has potential as an efficient AI-aided flanking sequence optimization approach for promoter design that may be of great benefit in synthetic biology applications. Availabilityhttps://github.com/WangLabTHU/deepseed

bioengineering↗

bioRxiv2023

EGFR inhibition in lung adenocarcinoma upregulates cell surface expression of the placental antigen ALPP and enhances efficacy of ALPP-ADC therapy

Alkaline phosphatase placental type (ALPP) and ALPPL2 are closely related and regulated GPI anchored proteins that are known to be expressed on the cell surface in some cancers, whereas normal tissue expression is largely limited to the placenta. Clinical utility of ALPP is potentially limited by heterogenous expression in tumors. Here, we assessed ALPP and ALPPL2 surfaceome protein levels in 158 cancer cell lines and mRNA expression levels in 10,967 tumors representing 32 cancer types from The Cancer Genome Atlas (TCGA), which revealed ALPP, and to a lesser extent ALPPL2, to be variably expressed in several cancer types including lung adenocarcinoma (LUAD). Surface expression of ALPP was confirmed by tissue microarray analysis of 204 lung tumors. Using LUAD as a model system, we demonstrated that treatment with EGFR inhibitors, or induction of cancer cell quiescence via nutrient deprivation greatly enhanced ALPP surface expression. Mechanistic studies revealed that enhancement of surface ALPP expression in LUAD following gefitinib treatment was mediated through repression of EGFR signaling and activation of the transcription factor FoxO3a, which was identified as an upstream transcriptional regulator of ALPP. Using xenograft models of LUAD, we further demonstrated that gefitinib treatment upregulates surface expression of ALPP in LUAD cells but not in normal tissues. Combination therapy with gefitinib and an ALPP antibody conjugated with Monomethylauristatin F (ALPP-ADC-MAF) resulted in superior anti-cancer efficacy compared with gefitinib or ALPP-ADC-MAF alone. Our findings support a novel combination treatment modality that boosts the efficacy of ALPP-ADC directed therapy.

cancer biology↗

bioRxiv2023

Guide-specific loss of efficiency and off-target reduction with Cas9 variants

High-fidelity Cas9 variants have been developed to reduce the off-target effects of CRISPR systems at a cost of efficiency loss. To systematically evaluate the efficiency and off-target tolerance of Cas9 variants in complex with different single guide RNAs (sgRNAs), we applied high-throughput viability screens and a synthetic paired sgRNA-target system to assess thousands of sgRNAs in combination with two high-fidelity Cas9 variants HiFi and LZ3. Comparing these variants against WT SpCas9, we found that ~20% of sgRNAs are associated with a significant loss of efficiency when complexed with either HiFi or LZ3. The loss of efficiency is dependent on the sequence context in the seed region of sgRNAs, as well as at positions 15-18 in the non-seed region that interacts with the REC3 domain of Cas9, suggesting that the variant-specific mutations in REC3 domain account for the loss of efficiency. We also observed various degrees of sequencedependent off-target reduction when different sgRNAs are used in combination with the variants. Given these observations, we developed GuideVar, a transfer-learning-based computational framework for the prediction of on-target efficiency and off-target effect with high-fidelity variants. GuideVar facilitates the prioritization of sgRNAs in the applications with HiFi and LZ3, as demonstrated by the improvement of signal-to-noise ratios in high-throughput viability screens using these high-fidelity variants.

systems biology↗

bioRxiv2023

An interpretable ML model to characterize patient-specific HLA-I antigen presentation

Personalized immunotherapy holds the promise of revolutionizing cancer prevention and treatment. However, selecting HLA-bound peptide targets that are specific to patient tumors has been challenging due to a lack of patient-specific antigen presentation models. Here, we present epiNB, a white-box, positive-example-only, semi-supervised method based on Naive Bayes formulation, with information content-based feature selection, to achieve accurate modeling using Mass Spectrometry data eluted from mono-allelic cell lines and patient-derived cell lines. In addition to achieving state-of-the-art accuracy, epiNB yields novel insights into the structural properties, such as interactions of peptide positions, that appear important for modeling personalized, tumor-specific antigen presentation. epiNB uses substantially less parameters than neural networks, does not require hyperparameter tweaking and can efficiently train and run on our web portal (https://epinbweb.streamlit.app/) or a regular PC/laptop, making it easily applicable in translational settings.

bioinformatics↗