Biology subjects

Xiao, A.

Publications and source records attributed to Xiao, A..

At least 19 recordsLinked to original sources

bioRxiv2026

Cross Study Regulator Activity Analysis Reveals Convergent Programs in alpha Synuclein PFF Models of Parkinson Disease

Alpha-synuclein preformed fibril (PFF) models reproduce features of Parkinson disease (PD), but differences in brain region, cell composition, and experimental design complicate comparisons across studies. We reconstructed study-specific regulatory networks using ARACNe (Algorithm for the Reconstruction of Accurate Cellular Networks) and inferred transcription factor (TF) and transcription cofactor activity using VIPER (Virtual Inference of Protein-activity by Enriched Regulon analysis). Five transcriptomic comparisons from four mouse PFF datasets were analyzed, with human PD single-nucleus RNA-sequencing data used to assess cross-species agreement. In the dopamine-neuron-enriched dataset GSE252967, PFF exposure was associated with increased activity of stress- and inflammation-related TFs, including JUND and RELA, and decreased activity of dopaminergic-identity regulators, including NR4A2, LMX1B, FOXA1, FOXA2, PITX3, and EN1. Several regulators were connected to Th in the inferred network. The substantia nigra bulk dataset GSE296779 showed reduced activity of ATF4- and XBP1-associated protein homeostasis and unfolded protein response programs, supported by regulon-weighted gene-set enrichment. Supporting mouse comparisons showed limited overlap in individual TFs but identified changes in inflammatory, glial, and stress-related programs. Agreement between inferred activity and transcript-level changes ranged from 64% to 100% in these comparisons. Human PD data showed cell-type-dependent agreement with mouse findings. EN1 and PITX3 were shared between the mouse and human networks upstream of Th/TH and showed reduced inferred activity in all seven human cell types, alongside nonsignificant decreases in the mouse dataset. These analyses identified changes in dopaminergic-identity, protein homeostasis, and inflammatory transcriptional programs across mouse PFF models, with partial support in human PD.

neuroscience↗

bioRxiv2026

Modeling human embryonic adrenogenesis in pluripotentstem cell-derived corticomedullar-like assembloids

During embryogenesis, the cortical and medullary compartments of the adrenal gland emerge from two distinct sources: intermediate mesoderm-derived steroidogenic progenitors that comprise the cortex, and migratory neural crest-derived catecholaminergic progenitors that give rise to the medulla. The distinct cellular origins and paracrine signaling environments that foster induction, differentiation, and/or migration of cortical and medullary layers present challenges to recapitulating their co-emergence and incorporation in a human pluripotent stem cell (hPSC) based system. To address this, we developed an approach for modeling the adreno-gonadal primordium (AGP) in compound organoids that are organized along a gradient of NR5A1 expression and which unexpectedly incorporate SOX10-expressing Schwann cell precursors that ultimately generate chromaffin cells. Deconstruction of paracrine signaling within AGP-like organoids (AGPLOs) identified a fibroblast growth factor 9 (FGF9)-driven mechanism that expands bridge cell-like chromaffin progenitors and correlates to paracrine signaling cues encountered by physiological correlates along their migration corridor to the cortex in vivo. Combination of NR5A1-expressing cells with chromaffin cell clusters generated corticomedullar-like assembloids (CMLAs) that generated steroids and catecholamines and were responsive to adrenocorticotropic hormone. This work establishes a platform for modeling the reciprocal signaling relationships that drive adrenal gland development and function in health and disease and provides an approach for generating hPSC-derived adrenal cells/tissues that can be applied therapeutically.

developmental biology↗

bioRxiv2026

Application of large language models to the annotation of cell lines and mouse strains in genomics data

Accurate, consistent and comprehensive metadata are essential for the reuse of functional genomics data deposited in repositories such as the Gene Expression Omnibus (GEO), however, achieving this often requires careful manual curation that is time-consuming, costly and prone to errors. In this paper, we evaluate the performance of Large Language Models (LLMs), specifically OpenAIs GPT-4o, as an assistive tool for entity-to-ontology annotation of two commonly encountered descriptors in transcriptomic experiments - mouse strains and cell lines. Using over 9,000 manually curated experiments from the Gemma database and over 5,000 associated journal articles, we assess the models ability to identify relevant free-text entries and map them to appropriate ontology terms. Using zero-shot prompting and retrieval-augmented generation (RAG) to incorporate domain-specific ontology knowledge, GPT-4o correctly annotated 77% of mouse strain and 59% of cell line experiments, and uncovered manual curation errors in Gemma for over 200 experiments. GPT-4o substantially outperformed a regular expression-based string-matching method, which correctly annotated only 6% of mouse strain experiments due to low precision. Model errors often arose from typographical mistakes or inconsistent naming in the GEO record or publication, and resembled those made by human curators. Along with annotations, our approach requests that the model output supporting context and quotes from the sources. These were typically accurate and enabled rapid curator verification. These findings suggest that LLMs are not ready to fully replace manual curators, but can already effectively support them. A human-in-the-loop workflow, in which LLMs annotations are provided to human curators for validation, may improve the efficiency and quality of large-scale biomedical metadata curation.

bioinformatics↗

bioRxiv2025

Stereoselective Degradation of Diacylglycerol Kinases Potentiate T cell Activation and Tumor Cell Cytotoxicity

Stereoselective recognition is a powerful means to differentiate selective versus non-specific activity of small molecules in complex biological systems. Here, we disclose stereochemically defined, sulfonyl-triazole inhibitors of the lipid enzyme diacylglycerol kinase-alpha (DGK), a key metabolic checkpoint for T cell effector function. Acute treatment with the covalent DGK inhibitor AHL-7160 recruited endogenous DGK to the plasma membrane in a stereoselective and isozyme-specific manner. The membrane translocation activity of AHL-7160 correlated with blockade of cellular phosphatidic acid production and potentiation of primary T cell-mediated killing of a glioblastoma cell line. Quantitative chemoproteomics revealed Y669 and K411 as sites of AHL-7160 modification on endogenous DGK in cells. Extended treatments resulted in proteasome-dependent and proteome-wide selective degradation of DGK in T cells. Collectively, these findings establish covalent DGK ligands as potent molecular glues with translational potential in immunotherapy.

biochemistry↗

bioRxiv2025

S1PR3 mediates glial stimulated tumor invasion in response to interstitial fluid flow

Cellular invasion is a primary challenge to complete resection and treatment of glioblastoma, the most aggressive and deadly primary brain tumor. The brain tumor microenvironment actively stimulates glioma invasion through a multitude of cellular, chemical, and biophysical cues. We and others have shown elevated interstitial fluid flow at the tumor border is one such biophysical cue that directly stimulates invasion through tumor-intrinsic signaling and, in other tumor types, priming of cancer-associated stromal cells. It is currently unclear if interstitial flow similarly primes neuroglial cells to promote glioma cell dissemination and can be targeted for therapeutic purposes. Here, we show elevated interstitial flow upregulates expression of sphingosine-1-phosphate receptor 3 (S1PR3) in glial astrocytes and microglia, which drives glioma cell invasion via chemotaxis. Flow-induced expression of glial S1PR3 is tumor-independent and displays a biphasic relationship to fluid shear stress magnitude in vitro and flow rate in vivo. Inhibition of glial S1PR3 in a tissue engineered culture model and orthotopic mouse model abrogates flow-stimulated invasion, demonstrating a tumor-extrinsic approach to limiting glioblastoma progression. Given prior evidence of a pro-inflammatory role for glial S1PR3, identification of S1PR3 as a disease-agnostic marker of flow-stimulated glia may also have therapeutic implications across myriad neuropathologies.

cancer biology↗

bioRxiv2025

Spatial Conformation of Pancreatic Cancer Is Associated with Disease Recurrence after Curative-Intent Total Neoadjuvant Therapy

Pancreatic ductal adenocarcinoma (PDAC) frequently recurs after total neoadjuvant therapy (TNT) and curative-intent resection. Traditional histopathologic response assessments demonstrate that major pathologic response (<5-10% viable residual cancer) is associated with favorable outcomes. However, most TNT cases achieve only a minor response. We investigated whether the spatial organization of residual PDAC encodes clinically meaningful biology beyond residual tumor burden. In a retrospective cohort of 203 resected PDAC patients, all with minor pathologic response, H&E whole-slide images were segmented into cancer and stroma to subsequently quantify spatial composition (e.g., patch size/density, edge density) and configuration (e.g., patch complexity/compactness, spatial intermixing) to model treatment-resistant tumor architecture. Non-response was associated with a more fragmented interface-rich ecology, higher edge density and diversity, and reduced homotypic aggregation, independent of conventional clinicopathologic features. This demonstrates a link between emergent tissue architecture of treated PDAC and therapeutic resistance. Further, two multivariable, spatial risk models were independently associated with disease-free survival (DFS): (1) cancer mean shape index and stromal shape-index variability (high-risk median DFS 7.23 vs 11.57 months; adjusted HR 1.75, p=0.007) and (2) mean stromal area with edge density (high- vs low-risk adjusted HR 1.94, p=0.006), outperforming traditional treatment response assessments. Quantifying residual cancer-stroma topology thus yields independent, prognostic signals in post-TNT PDAC and motivates prospective, spatially informed adjuvant strategies and mechanistic studies of edge habitats and mixing as therapeutic vulnerabilities.

cancer biology↗

bioRxiv2025

Non-disruptive 3D profiling of combinations of epigenetic marks in single cells

Recent advancements in single-cell sequencing and spatial omics technologies have enhanced our understanding of diverse cellular identities, compositions, architectures, and functions. However, the single-cell three-dimensional (3D) organization of the epigenome is still not well understood, due to an absence of spatial single-cell methods that allow high-resolution, locus-specific detection of combinations of epigenetic marks while maintaining the 3D organization of the genome. Here, we develop Epigenetic Proximity Hybridization Reaction (Epi-PHR), a non-disruptive image-based single-cell spatial epigenetic profiling technology. Epi-PHR enables locus-specific and high-resolution in situ detection of combinations of epigenetic marks at hundreds of single gene targets within the same individual cells, while retaining the 3D organization of the genome. Dual-mark Epi-PHR in hippocampus tissue sections revealed region-specific epigenetic profiles. Phased Epi-PHR combined with chromatin tracing simultaneously detects allele-specific epigenetic states and chromatin conformations of a paternally imprinted Meg3 gene cluster in single cells, revealing associations between specific epigenetic mark enrichment and chromatin folding features for the distinct alleles from different parental origins. Surprisingly, we found that H3K9me3 abundance is positively associated with a chromatin domain boundary at the Meg3 locus among maternal copies. We expect Epi-PHR will be broadly applicable in research requiring single-cell spatial epigenetic information, and will help extend our understanding of the combinatorial epigenetic code and its relationship with chromatin organization.

genetics↗

bioRxiv2025

Epigenetic priming promotes acquisition of tyrosine kinase inhibitor resistance and oncogene amplification in human lung cancer

In mammalian cells, gene copy number is tightly controlled to maintain gene expression and genome stability. However, a common molecular feature across cancer types is oncogene amplification, which promotes cancer progression by drastically increasing the copy number and expression of tumor-promoting genes. For example, in tyrosine kinase inhibitor (TKI)-resistant lung adenocarcinoma (LUAD), oncogene amplification occurs in over 40% of patients tumors. Despite the prevalence of oncogene amplification in TKI-resistant tumors, the mechanisms facilitating oncogene amplification are not fully understood. Here, we find that LUADs exhibit a unique chromatin signature demarcated by strong CTCF and cohesin deposition in drug-naive tumors, which correlates with the boundaries of oncogene amplicons in TKI-resistant LUAD cells. We identified a global chromatin priming effect during the acquisition of TKI resistance, marked by a dynamic increase of H3K27Ac, cohesin loading, and inter-TAD interactions, which occurs before the onset of oncogene amplification. Furthermore, we have found that the METTL7A protein, which was previously reported to localize to the endoplasmic reticulum and inner nuclear membrane, has a novel chromatin regulatory function by binding to amplified loci and regulating cohesin recruitment and inter-TAD interactions. Surprisingly, we discovered that METTL7A remodels the chromatin landscape prior to large-scale copy number gains. Furthermore, while METTL7A depletion has little effect on the chromatin structure and proliferation of drug-naive cells, METTL7A depletion prevents the formation and maintenance of TKI resistant-clones, highlighting the specific role of METTL7A as cells are becoming resistant. In summary, we discovered an unexpected mechanism required for the acquisition of TKI resistance regulated by a largely uncharacterized factor, METTL7A. This discovery sheds light into the maintenance of oncogene copy number and paves the way to the development of new therapeutics for preventing TKI resistance in LUAD.

genetics↗

bioRxiv2024

METTL7A improves bovine IVF embryo competence by attenuating oxidative stress

In vitro fertilization (IVF) is a widely used assisted reproductive technology to achieve a successful pregnancy. However, the acquisition of oxidative stress in embryo in vitro culture impairs its competence. Here, we demonstrated that a nuclear coding gene, methyltransferase- like protein 7A (METTL7A), improves the developmental potential of bovine embryos. We found that exogenous METTL7A modulates expression of genes involved in embryonic cell mitochondrial pathways and promotes trophectoderm development. Surprisingly, we discovered that METTL7A alleviates mitochondrial stress and DNA damage and promotes cell cycle progression during embryo cleavage. In summary, we have identified a novel mitochondria stress eliminating mechanism regulated by METTL7A that occurs during the acquisition of oxidative stress in embryo in vitro culture. This discovery lays the groundwork for the development of METTL7A as a promising therapeutic target for IVF embryo competence. Summary statement (Graphic abstract)We describe a molecule acts in the pre-implantation period to attenuate oxidative stress that enhances embryo development to the blastocyst stage and subsequent pregnancy in cattle. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=163 SRC="FIGDIR/small/628915v1_ufig1.gif" ALT="Figure 1"> View larger version (43K): org.highwire.dtl.DTLVardef@12863c7org.highwire.dtl.DTLVardef@aec66aorg.highwire.dtl.DTLVardef@648ccaorg.highwire.dtl.DTLVardef@1572927_HPS_FORMAT_FIGEXP M_FIG C_FIG

developmental biology↗

bioRxiv2023

Single-cell transcriptome analysis reveals the cellular atlas of human intracranial aneurysm and highlights inflammation features associated with aneurysm rupture

Intracranial aneurysm (IA) is pouch-like pathological dilations of cerebral arteries, which often affects middle-aged people and culminates in life-threatening hemorrhagic stroke. A deeper knowledge of the cellular and gene expression perturbations in human IA tissue deepens our understanding of disease mechanisms and facilitates developing pharmacological targets for unruptured IA. In this study, 21,332 qualified cells were obtained from cell-sparse ruptured and unruptured human IA tissues and a detailed cellular profile was determined, including conventional endothelial cells, smooth muscle cells (SMC), fibroblasts and the newly identified pericytes. Notably, striking proportion of immune cells were identified in IA tissue, with the number of monocyte/macrophages and neutrophils being remarkably higher in ruptured IA. By leveraging external datasets and machine learning algorithms, a subset of macrophages characterized by high expression of CCL3 and CXCL3, and transcriptional activation of NF-{kappa}B and HIVEP2 was identified as the cell most associated with IA rupture. Further, the interactome of CCL3/CXCL3 macrophages disclosed their role in regulating vascular cell survival and orchestrating inflammation. In summary, this study illustrated the profile and interactions of vascular and immune cells in human IA tissue and the opportunities for targeting local chronic inflammation.

cell biology↗

bioRxiv2022

The ribosome quality control factors Asc1 and Hel2 regulate the expression of HSP70 during heat shock and recovery

Cells rapidly adapt to survive harsh environmental conditions through the potent upregulation of molecular chaperones or heat shock proteins (HSPs). The inducible members of the HSP70 family are the fastest and most transcriptionally induced chaperone upon stress. The HSP70 mRNA life cycle regulation in the cytoplasm is unique because it is translated during stress when general translation is repressed and rapidly degraded once conditions are optimal for growth. Contrary to the role of the HSP70 mRNA 5 untranslated region in maximizing the synthesis of HSP70, we discovered that the coding sequence (CDS) represses its translation through the ribosome quality control (RQC) mechanism. The CDS of the most inducible HSP70 in Saccharomyces cerevisiae, SSA4, is uniquely biased with low-frequency codons that promote ribosome stalling during heat stress. The stalled ribosomes are recognized by RQC components Asc1p and Hel2p and two ribosome proteins, Rps28A and Rps19B, that we identified as new RQC components. Surprisingly, RQC does not signal the degradation of the SSA4 mRNA by no-go-decay (NGD). Instead, Asc1p destabilizes the SSA4 mRNA during recovery from heat stress by a mechanism independent of its ribosome binding and SSA4 CDS codon optimality. Therefore, Asc1p operates two synergistic mechanisms that converge to regulate the life cycle of HSP70 mRNA during stress and recovery. Our research identifies Asc1p as a critical regulator of the stress response and RQC as the system tuning HSP70 synthesis.

molecular biology↗

bioRxiv2021

Double-check the zebrafish 18s rRNA qPCR primers: they may be wrong

A widely used qPCR primer for zebrafish gene rna18s (18s rRNA, or 18s), with the sequence of 5-TCGCtaGTtGGCATCGTTTAtG-3, is found to be incorrect. Initially designed for rainbow trout (Oncorhynchus mykiss) rna18s, the primer has four different nucleotides from the zebrafish sequence 5-TCGCGGGTCGGCATCGTTTACG-3 (indicated in bold/underlined, lowercase letters for rainbow trout and uppercase letters for zebrafish). Since its first use in zebrafish in 2006, this mismatched primer has been clearly stated to be used in at least 50 publications and may have affected hundreds or more in publications citing them. For a sensitive, quantitative method as qPCR, this error must be corrected as soon as possible in the zebrafish community by using rna18s primer sets with accurate sequences, such as those summarized and newly designed in this article.

molecular biology↗

bioRxiv2020

The CoREST Repressor Complex Mediates Phenotype Switching and Therapy Resistance in Melanoma

Virtually all patients with BRAF-mutant melanoma develop resistance to MAPK inhibitors largely through non-mutational events1,2. Although the epigenetic landscape has been shown to be altered in therapy-resistant melanomas and other cancers3,4, a specific targetable epigenetic mechanism regulating treatment resistance has not been validated to date. Here we evaluate the CoREST repressor complex and the novel inhibitor, corin5, within the context of melanoma phenotype plasticity and therapeutic resistance in order to define epigenetic mechanisms underlying these processes. We find that CoREST is a critical mediator of the major distinct melanoma phenotypes and that corin treatment of melanoma cells leads to phenotype reprogramming. We further demonstrate that treatment of BRAF inhibitor (BRAFi)-resistant melanomas with corin leads to resensitization of tumor cells to BRAFi. Among the transcriptional targets of CoREST in melanoma are the dual-specificity phosphatases (DUSPs). DUSP1 is shown to be consistently downregulated in BRAFi-resistant melanomas which can be reversed by corin treatment, thereby leading to downstream inhibition of p38 MAPK activity and resensitization of resistant cells to targeted BRAFi therapies. These findings identify the CoREST repressor complex as a central mediator of melanoma phenotype plasticity and resistance to targeted therapy and suggest that CoREST inhibitors may prove beneficial to patients with BRAF-mutant melanomas who have acquired BRAFi-resistance.

cancer biology↗

bioRxiv2019

Enhance genome editing efficiency and specificity by a quick CRISPR/Cas9 system

CRISPR/Cas9 is a powerful genome editing tool that has been successfully applied to a variety of species, including zebrafish. However, targeting efficiencies vary greatly at different genomic loci, the underlying causes of which were still elusive. Here we report a quick CRISPR/Cas9 system, designated as qCas9, which exhibits accelerated turnover of Cas9 protein in zebrafish. Our data showed that qCas9 significantly improved targeting efficiency, including both knock-out and knock-in in F0 embryos, and yielded higher germline transmission rate in founder screen. Importantly, qCas9 showed little to no off-target editing in zebrafish and profoundly reduced off-target effect in HEK293T cell line. In summary, our findings demonstrate that qCas9 is a simple, economic and highly effective method to improve genome editing efficiency in zebrafish embryos and also holds great potential in reducing off-target effect in mammalian cell lines.

genetics↗

bioRxiv2019

PRE-BORDER GENE FOXB1 REGULATES THE DIFFERENTIATION TIMING AND AUTONOMIC NEURONAL POTENTIAL OF HUMAN NEURAL CREST CELLS

What are the factors that are induced during the transitory phases from pluripotent stem cells to lineage specified cells, how are they regulated, and what are their functional contributions are fundamental questions for basic developmental biology and clinical research. Here, we uncover a set of pre-border (pB) gene candidates, including forkhead box B1 (FOXB1), induced during human neural crest (NC) cell development. We characterize their associated enhancers that are bound by pluripotency factors and rapidly activated by {beta}-catenin-mediated signaling during differentiation. Surprisingly, the endogenous transient expression of FOXB1 directly regulates multiple early NC and neural progenitor loci including PAX7, MSX2, SOX1, and ASCL1, controls the timing of NC fate acquisition, and differentially activates autonomic neurogenic versus mesenchymal fates in mature NC cells. Our findings provide further insight into the concept of the less characterized pB state and clearly establishes FOXB1 as a key regulator in early cell fate decisions during human pluripotent stem cell differentiation.

developmental biology↗

bioRxiv2018

Rationalizing Translation Elongation by Reinforcement Learning

Translation elongation plays a crucial role in multiple aspects of protein biogenesis. In this study, we develop a novel deep reinforcement learning based framework, named RiboRL, to model the distributions of ribosomes on transcripts. In particular, RiboRL employs a policy network (PolicyNet) to perform a context-dependent feature selection to facilitate the prediction of ribosome density. Extensive tests demonstrate that RiboRL can outperform other state-of-the-art methods in predicting ribosome densities. We also show that the reinforcement learning based strategy can generate more informative features for the prediction task when compared to other commonly used attribution methods in deep learning. Moreover, the in-depth analyses and a case study also indicate the potential applications of the RiboRL framework in generating meaningful biological insights regarding translation elongation dynamics. These results have established RiboRL as a useful computational tool to facilitate the studies of the underlying mechanisms of translational regulation.

bioinformatics↗

bioRxiv2018

NeoDTI: Neural integration of neighborinformation from a heterogeneous network fordiscovering new drug-target interactions

MotivationAccurately predicting drug-target interactions (DTIs) in silico can guide the drug discovery process and thus facilitate drug development. Computational approaches for DTI prediction that adopt the systems biology perspective generally exploit the rationale that the properties of drugs and targets can be characterized by their functional roles in biological networks.\n\nResultsInspired by recent advance of information passing and aggregation techniques that generalize the convolution neural networks (CNNs) to mine large-scale graph data and greatly improve the performance of many network-related prediction tasks, we develop a new nonlinear end-to-end learning model, called NeoDTI, that integrates diverse information from heterogeneous network data and automatically learns topology-preserving representations of drugs and targets to facilitate DTI prediction. The substantial prediction performance improvement over other state-of-the-art DTI prediction methods as well as several novel predicted DTIs with evidence supports from previous studies have demonstrated the superior predictive power of NeoDTI. In addition, NeoDTI is robust against a wide range of choices of hyperparameters and is ready to integrate more drug and target related information (e.g., compound-protein binding affinity data). All these results suggest that NeoDTI can offer a powerful and robust tool for drug development and drug repositioning.\n\nAvailability and implementationThe source code and data used in NeoDTI are available at: https://github.com/FangpingWan/NeoDTI.\n\nContactzengjy321@tsinghua.edu.cn\n\nSupplementary informationSupplementary data are available at Bioinformatics online.

systems biology↗

bioRxiv2018

DeepHINT: Understanding HIV-1 integration via deep learning with attention

MotivationHuman immunodeficiency virus type 1 (HIV-1) genome integration is closely related to clinical latency and viral rebound. In addition to human DNA sequences that directly interact with the integration machinery, the selection of HIV integration sites has also been shown to depend on the heterogeneous genomic context around a large region, which greatly hinders the prediction and mechanistic studies of HIV integration.\n\nResultsWe have developed an attention-based deep learning framework, named DeepHINT, to simultaneously provide accurate prediction of HIV integration sites and mechanistic explanations of the detected sites. Extensive tests on a high-density HIV integration site dataset showed that DeepHINT can outperform conventional modeling strategies by automatically learning the genomic context of HIV integration solely from primary DNA sequence information. Systematic analyses on diverse known factors of HIV integration further validated the biological relevance of the prediction result. More importantly, in-depth analyses of the attention values output by DeepHINT revealed intriguing mechanistic implications in the selection of HIV integration sites, including potential roles of several basic helix-loop-helix (bHLH) transcription factors and zinc-finger proteins. These results established DeepHINT as an effective and explainable deep learning framework for the prediction and mechanistic study of HIV integration.\n\nAvailabilityDeepHINT is available as an open-source software and can be downloaded from https://github.com/nonnerdling/DeepHINT\n\nContactlzhang20@mail.tsinghua.edu.cn and zengjy321@tsinghua.edu.cn

bioinformatics↗