Biology subjects

Su, X.

Publications and source records attributed to Su, X..

At least 91 records · Page 5Linked to original sources

bioRxiv2023

BRG1/BRM inhibitor targets AML stem cells and exerts superior preclinical efficacy combined with BET or Menin inhibitor

BRG1 (SMARCA4) and BRM (SMARCA2) are the core ATPases of chromatin remodeling BAF (BRG1/BRM-associated factor) complexes, which enable transcription factors/co-factors to modulate gene-expressions, mediating growth, differentiation-arrest and survival of AML cells. In AML with MLL1r (MLL1 rearrangement) or mutant (mt) NPM1, although monotherapy with Menin inhibitor (MI) induces clinical remissions, most patients either fail to respond or relapse. FHD-286 is a selective BRG1/BRM inhibitor, undergoing clinical development in AML. Here, FHD-286 induced differentiation and lethality in AML cells with MLL1r or mtNPM1, reducing chromatin accessibility and repressing c-Myc, PU.1 and CDK4/6. Whereas FHD-286 monotherapy reduced AML burden, leukemia-initiating potential and improved survival, FHD-286 combinations with MI, BET inhibitor, decitabine or venetoclax was significantly more effective in reducing AML burden and improved survival, without significant toxicity, in xenograft models of AML with MLL1r or mtNPM1. These findings highlight promising FHD-286-based combinations for therapy of AML with MLL1r or mtNPM1. Statement of SignificanceInhibition of BRG1/BRM ATPases by FHD-286 reduced chromatin accessibility, repressed c-Myc, PU.1 and CDK4/6, inducing differentiation, leukemia-initiating potential and lethality in AML stem-progenitor cells. FHD-286-based combinations with Menin or BET inhibitor or decitabine reduced AML burden and improved survival in xenograft models of AML with MLL rearrangement or mutant NPM1.

cancer biology↗

bioRxiv2023

Dynamics of endogenous and water cortisol release in Asian Seabass Lates calcarifer after acute stress in a farm scale recirculating aquaculture system

Stress in farmed fish is associated with poor feeding, slow growth, disease, and mortality. Therefore, it is essential to closely monitor the stress levels in fish to optimize farming practices which could then enhance productivity and welfare in aquaculture operations. Cortisol, a stress hormone that can be found in the blood, is a reliable biomarker for evaluating fish stress. As blood sampling is highly invasive, alternative cortisol sampling methods such as fin, mucus, and the surrounding water which contains released cortisol, have been proposed as less invasive or non-invasive sampling methods. However, a comprehensive understanding of their temporal dynamics and associations with plasma cortisol levels is still lacking. In this study, we subjected Lates calcarifer, Asian sea bass within a farm-scale (3,000 L tank, 9,000 L system) high-flow rate (8,000 L/hour) Recirculating Aquaculture System (RAS) to an acute handling stress challenge specifically involving chasing and air exposure, and quantified cortisol dynamics both within different biological samples including blood, fin, and mucus and in tank water from multiple sampling points. We showed that handling stress induced an expected increase in plasma and mucosal cortisol, peaking at 1 hour and 24-48 hours, respectively, and that plasma and mucus cortisol were moderately correlated, especially during the stress period. Fin cortisol did not show consistent dynamics. Water cortisol similarly rose, but peaked within 40 minutes from the start of the stressor, in a pattern that was dependent on the site of sampling within the RAS system, likely due to RAS circulation dynamics. Our study is the first to examine the impact of stress on cortisol accumulation and release in Asian Sea bass in a farm-scale RAS, thus complementing existing research on the efficacy of fin, mucus, and water cortisol as stress indicators that could help optimize aquaculture productivity and welfare.

physiology↗

bioRxiv2023

KaScape: A sequencing-based method for global characterization of protein-DNA binding affinity

It is difficult to exhaustively screen all possible DNA binding sequences for a given transcription factor (TF). Here, we develop a method named "KaScape", by which, TFs bind to all possible DNA sequences in the same DNA pool where the DNA sequences are prepared by randomized oligo synthesis and the random length can be adjusted to, e.g., 4, 5, 6, or 7, etc. After separating the bound from unbound double-strand DNA, their sequences are determined by next-generation sequencing. To demonstrate the relative binding affinities of all possible DNA sequences determined by KaScape, we develop a three-dimensional KaScape viewing software based on a K-mer graph. We apply KaScape to 12 plant TF family AtWRKY proteins and find that all AtWRKY proteins bind to the core sequence GAC with similar profiles. KaScape not only can detect binding sequences that are consistent with the consensus W-box "TTGAC(C/T)", but also other sequences with weak affinity. KaScape provides a high-throughput, easy-to-operate, sensitive, and exhaustive method to quantitatively characterize the relative binding strength of a TF to all possible binding sequences, allowing us to comprehensively characterize the specificity and affinity landscape of transcription factors, particularly for moderate and low affinity binding sites. HighlightsO_LIA general method, KaScape, using NGS (next-generation sequencing) with a series of random dsDNA for exhaustive characterization of protein-DNA binding. C_LIO_LIA K-mer-based analysis and display software tool KGViewer developed to analyze the relative affinity landscape of KaScape results. C_LI

biochemistry↗

bioRxiv2023

Use of Artificial Intelligence for Automated Detection and Surveillance of Red Imported Fire Ants Nests

The red imported fire ant, Solenopsis invicta Buren is a destructive invasive species that has spread around the world. Early detection of S. invicta nests is critical for effective monitoring and control in invaded regions. This study presents a novel surveillance system for S. invicta nests combining artificial intelligence and robotic dogs. The system was designed with intelligent recognition algorithms to accurately identify S. invicta nests. With a precision rate of 95%, the system yielded efficient detection of S. invicta nests with increased sensitivity and low missing and false discovery rates, which can aid or even replace humans in locating and delivering pesticides to fire ant nests in open fields to effective control of this invasive pest.

ecology↗

bioRxiv2023

Deep learning and host variable embedding augment microbiome-based simultaneous detection of multiple diseases

Microbiome has emerged as a promising indicator or predictor of human diseases. However, previous studies typically labeled each specimen as either healthy or with a specific disease, ignoring the prevalence of complications or comorbidities in actual cohorts, which may confound the microbial-disease associations. For instance, a patient may suffer from multiple diseases, making it challenging to detect their health status accurately. Furthermore, host phenotypes such as physiological characteristics and lifestyles can alter the microbiome structure, but this information has not yet been fully utilized in data models. To address these issues, we propose a highly explainable deep learning (DL) method called Meta-Spec. Using a deep neural network (DNN) based approach, it encodes and embeds the refined host variables with microbiome features, enabling the detection of multiple diseases and their correlations simultaneously. Our experiments showed that Meta-Spec outperforms regular machine learning (ML) strategies for multi-label disease screening in several cohorts. More importantly, Meta-Spec can successfully detect comorbidities that are often missed by regular ML approaches. In addition, due to its high interpretability, Meta-Spec captures key factors that shape disease patterns from host variables and microbial members. Hence, these efforts improve the feasibility and sensitivity of microbiome-based disease screening in practical scenarios, representing a significant step towards personalized medicine and better health outcomes.

bioinformatics↗

bioRxiv2023

Spatially-resolved single-cell atlas of ascidian endostyle provides insights into the origin of vertebrate pharyngeal organs

The pharynx is an endoderm innovation in deuterostome ancestors, the vertebrate descendent structure of which is a pharyngeal developmental organizer involving multi-germ layer and organ derivatives. However, the evolutionary origination of complicated pharynx organs in vertebrates is still largely unknown. Endostyle, a transitional pharyngeal organ exclusively in basal chordates provides an opportunity to reveal the origin of pharyngeal organs. Here, utilizing cutting-edged Stereo-seq and single-cell RNA-seq, we constructed the first spatially-resolved single-cell atlas in the endostyle of urochordate ascidian Styela clava, where the spatial location of Stereo-seq and high capture efficiency of single-cell RNA-seq complement each other and identified 23 highly differentiated cell types. We identified a previously overlooked hemolymphoid region (HLR), which harbors immune and blood cell clusters with enriched stemness capacities, illuminating a mixed rudiment and stem-cell niches for the blood and lymphoid system. More excitingly, we discovered a mechanical-sensitive hair cell candidate in zone 3 homologous to vertebrate acoustico-lateralis system, which was supported by the expression of in situ hybridization-verified inner ear-specific markers, including PTPRQ, USH2A, WHRN, and ADGRV1, ultracellular structure evidence and cross-species comparison. These results thoroughly renewed the comprehension of the basal-chordate pharynx and provides expressional evidence for multiplexed pharyngeal organ evolution.

evolutionary biology↗

bioRxiv2023

Therapeutic targeting of ACLY in T-ALL in vivo

T-cell Acute Lymphoblastic Leukemia (T-ALL) is a hematological malignancy in need of novel therapeutic approaches. Here, we identify the ATP-citrate lyase ACLY as overexpressed and as a novel therapeutic target in T-ALL. To test the effects of ACLY in leukemia progression, we developed an isogenic model of NOTCH1-induced Acly conditional knockout leukemia. Importantly, we observed intrinsic antileukemic effects upon loss of ACLY, which further synergized with NOTCH1 inhibition in vivo. Metabolomic profiling upon ACLY loss revealed a metabolic crisis with reduced acetyl-CoA levels, as well as a decreased oxygen consumption rate. Gene expression profiling analyses showed that the transcriptional signature of ACLY loss very significantly correlates with the signature of MYC loss in vivo. Mechanistically, the decrease in acetyl-CoA led to reduced H3K27ac levels in Myc, resulting in transcriptional downregulation of Myc and drastically reduced MYC protein levels. Interestingly, our analyses also revealed a reciprocal relationship whereby ACLY itself is a direct transcriptional target of MYC, thus establishing a feedforward loop that is important for leukemia progression. Overall, our results identified a relevant ACLY-MYC axis and unveiled ACLY as a novel promising target for T-ALL treatment.

cancer biology↗

bioRxiv2023

A novel order-level lineage of ammonia-oxidizing Thaumarchaeota is widespread in marine and terrestrial environments

Ammonia-oxidizing archaea (AOA) are among the most ubiquitous and abundant groups of Archaea on Earth, widely distributed in marine, terrestrial, and geothermal ecosystems. However, the genomic diversity, biogeography, and evolutionary process of AOA populations in subsurface environments are vastly understudied compared to those of marine and soil AOA. We here report a novel AOA order Candidatus Nitrosomirales that forms a deeply branching basal sister lineage to the thermophilic Ca. Nitrosocaldales. Metagenomic and 16S rRNA gene read mapping demonstrates the dominant presence of Nitrosomirales AOA in various groundwater environments and their widespread distribution across a range of geothermal, terrestrial, and marine habitats. Notably, terrestrial Nitrosomirales AOA show the genetic capacity of using formate as an alternative source of reductant and appear to have acquired key metabolic genes and operons from other mesophilic populations via horizontal gene transfer, including the genes encoding urease, nitrite reductase, and V-type ATPase. Potential metabolic versatility and acquired functions may facilitate their radiation into a variety of subsurface, marine, and soil environments. Molecular thermometer-based evolutionary analysis suggests that Nitrosomirales originated from thermophilic environments and transitioned into temperate habitats in parallel with Nitrososphaerales and Nitrosopumilales. We also provide evidence that terrestrial-marine habitat transitions occurred within each one of the four AOA orders, which reveals a more complex evolutionary trajectory of major AOA lineages than previously proposed. Together, these findings establish a robust taxonomic and evolutionary framework of AOA and provide new insights into the ecology and evolution of this globally abundant functional guild.

microbiology↗

bioRxiv2023

Linking genotypic and phenotypic changes in the LTEE using metabolomics

1Changes in an organisms environment, genome, or gene expression patterns can lead to changes in its metabolism. The metabolic phenotype can be under selection and contributes to adaptation. However, the networked and convoluted nature of an organisms metabolism makes relating mutations, metabolic changes, and effects on fitness challenging. To overcome this challenge, we use the Long-Term Evolution Experiment (LTEE) with E. coli as a model to understand how mutations can eventually affect metabolism and perhaps fitness. We used mass-spectrometry to broadly survey the metabolomes of the ancestral strains and all 12 evolved lines. We combined this metabolic data with mutation and expression data to suggest how mutations that alter specific reaction pathways, such as the biosynthesis of nicotinamide adenine dinucleotide, might increase fitness in the system. Our work provides a better understanding of how mutations might affect fitness through the metabolic changes in the LTEE and thus provides a major step in developing a complete genotype-phenotype map for this experimental system.

evolutionary biology↗

bioRxiv2023

Analysis of Biodeterioration on Ancestral Halls with Different Visitor Flow Rates Using Amplicon Sequencing and Conventional Culturing Methods

AbstractWooden buildings are facing biodeterioration due to natural weathering. Ancestral halls are traditional Chinese wooden architecture with high artistic value. The research analyzed the microbial diversity of nine ancestral halls in Guangdong Province under subtropical monsoon climates. With amplicon sequencing, we found a common core microorganism group, including Bacillus, Pseudomonas, Paenibacillus, Acinetobacter, Toxicocladosporium, Cladosporium, Aspergillus, and Epicoccum. A total of 8 different species of bacteria and 7 different species of fungi were isolated from Chen Clan Ancestral halls, most of which were proven to be capable of degrading cellulose or lignin. The comparison between damaged and undamaged sites (whether the paint coating is flaked off) revealed similar microbial compositions but with differing abundances. Firmicutes and Ascomycota dominated damaged surfaces, while Proteobacteria and Actinobacteria prevailed in undamaged ones. Notably, Cladosporium, widely present in both sites, emerged as the primary cellulose-degrading genus. Given the similar microbial compositions between damaged and undamaged sites, it is suggested that using paint alone as the exclusive means to protect bare wood is insufficient. Apparent correlation between microorganisms on wooden surfaces and humans showed that the increase of visitor flow rates might impact the microbiota composition on cultural relics. We evaluated the effect of four representative biocides and determined the feasibility of low-concentration Isothiazolinone. This research facilitates a basic understanding of the microbial weathering of ancestral halls and add significant reference for preserving wooden structures.

microbiology↗

bioRxiv2023

At-home oral bacteria collection using a lollipop-based microfluidic device

Respiratory infections are common in children, and there is a need for user-friendly collection methods. Here, we performed the first human subjects study using the CandyCollect device, a lollipop inspired saliva collection device.1 We showed the CandyCollect device can be used to collect salivary bacteria from healthy adults using Streptococcus mutans and Staphylococcus aureus as proof-of-concept commensal bacteria. We enrolled healthy adults in a nationwide (USA) remote study in which participants were sent study packages containing CandyCollect devices and traditional commercially available oral swabs and spit tubes. Participants sampled themselves at home, completed usability and user preference surveys, and mailed the samples back to our laboratory for analysis by qPCR. Our results showed that for participants in which a given bacterium (S. mutans or S. aureus) was detected in one or both of the commercially available methods (oral swab and/or spit tubes), CandyCollect devices had a 100% concordance with the positive result (n=14 participants). Furthermore, the CandyCollect device was ranked the highest preference sampling method among the three sampling methods by 26 participants surveyed (combining survey results across two enrollment groups). We also showed that the CandyCollect device has a shelf life of up to 1 year at room temperature, a storage period that is convenient for clinics or patients to keep the CandyCollect device and use it any time. Taken together, we have demonstrated that the CandyCollect is a user-friendly saliva collection tool that has the potential to be incorporated into diagnostic assays in clinic visits and telemedicine. For Table of Contents Only O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=126 SRC="FIGDIR/small/524039v6_ufig1.gif" ALT="Figure 1"> View larger version (37K): org.highwire.dtl.DTLVardef@1356b32org.highwire.dtl.DTLVardef@10c737org.highwire.dtl.DTLVardef@1872455org.highwire.dtl.DTLVardef@1625069_HPS_FORMAT_FIGEXP M_FIG C_FIG

biochemistry↗

bioRxiv2022

DeepAIR: a deep-learning framework for effective integration of sequence and 3D structure to enable adaptive immune receptor analysis

Structural docking between the adaptive immune receptors (AIRs), including T cell receptors (TCRs) and B cell receptors (BCRs), and their cognate antigens is one of the most fundamental processes in adaptive immunity. However, current methods for predicting AIR-antigen binding largely rely on sequence-derived features of AIRs, omitting the structure features that are essential for binding affinity. In this study, we present a deep-learning framework, termed DeepAIR, for the accurate prediction of AIR-antigen binding by integrating both sequence and structure features of AIRs. DeepAIR consists of three feature encoders (a trainable-embedding-layer-based gene encoder, a transformer-based sequence encoder, and a pre-trained AlphaFold2-based structure encoder), a gating-based attention mechanism to extract important features, and a tensor fusion mechanism to integrate obtained features. We train and evaluate DeepAIR on three downstream prediction tasks, including the prediction of AIR-antigen binding affinity, the prediction of AIR-antigen binding reactivity, and the classification of the immune repertoire. On five representative datasets, DeepAIR shows outstanding prediction performance in terms of AUC (area under the ROC curve) in predicting the binding reactivity to various antigens, as well as the classification of immune repertoire for nasopharyngeal carcinoma (NPC) and inflammatory bowel disease (IBD). DeepAIR is freely available for academic purposes at https://github.com/TencentAILabHealthcare/DeepAIR. We anticipate that DeepAIR can serve as a useful tool for characterizing and profiling antigen binding AIRs, thereby informing the design of personalized immunotherapy. HighlightsO_LIIntegrating predicted AIR structures using AlphaFold2 significantly improves the prediction accuracy of the binding reactivity between AIRs and antigens. C_LIO_LIDeepAIR is featured by a novel deep learning architecture that leverages both the gating-based attention mechanism and tensor fusion mechanism to effectively extract and integrate informative features from three feature encoders, including a trainable embedding-layer-based gene encoder, a transformer-based sequence encoder, and a pre-trained AlphaFold2-based structure encoder. C_LIO_LIDeepAIR is implemented as a biologically interpretable deep learning framework that highlights the key residues in both and {beta} chains that are critical for predicting the AIR-antigen binding. C_LI

immunology↗

bioRxiv2022

The Slingshot phosphatase 2 is required for acrosome biogenesis during spermatogenesis in mice

The acrosome is a membranous organelle positioned in the anterior portion of sperm head and is essential for male fertility. Acrosome biogenesis requires the dynamic cytoskeletal shuttling of vesicles towards nascent acrosome which is regulated by a series of accessory proteins. However, much remains unknown about the molecular basis underlying this process. Here, we generated Ssh2 knock-out (KO) mice and show that Slingshot phosphatase 2 (SSH2), a regulator of actin remodeling, is essential for acrosome biogenesis and male fertility. In Ssh2 KO males, spermatogenesis was arrested at the early spermatid stage with enhanced germ cell apoptosis and the impaired acrosome biogenesis was characterized by defective transport/fusion of proacrosomal vesicles. Moreover, disorganized F-actin structures accompanied by excessive phosphorylation of COFILIN were observed in testes of Ssh2 KO mice. Collectively, our data reveal a modulatory role for SSH2 in acrosome biogenesis through COFILIN-mediated actin remodeling and the indispensability of this phosphatase in male fertility in mice.

developmental biology↗

bioRxiv2022

A geometric deep learning framework for drug repositioning over heterogeneous information networks

The effectiveness of computational drug repositioning techniques has been further improved due to the development of artificial intelligence technology. However, most of the existing approaches fall short of taking into account the non-Euclidean nature of biomedical data. To overcome this problem, we propose a geometric deep learning (GDL) framework, namely DDAGDL, to predict drug-disease associations (DDAs) on heterogeneous information networks (HINs). DDAGDL can take advantage of complicated biological information to learn the feature representations of drugs and diseases by ingeniously projecting drugs and diseases including geometric prior knowledge of network structure in a non-Euclidean domain onto a latent feature space. Experimental results show that DDAGDL is able to identify high-quality candidates for Alzheimers disease (AD) and Breast neoplasms (BN) that have already been reported by previously published studies, and some of them are not even identified by comparing models.

bioinformatics↗

bioRxiv2022

Tumor subclones, where are you?

IntroductionTumor clonal structure is closely related to future progression, which has been mainly investigated via mutation abundance clustering in bulk sample. With limited studies at single-cell resolution, a systematic comparison of the two approaches is still lacking. MethodsHere, using bulk and single-cell mutational data from liver and colorectal cancers, we would like to check the possibility of obtaining accurate tumor clonal structures from bulk-level analysis. We checked whether co-mutations determined by single-cell analysis had corresponding bulk variant allele frequency (VAF) peaks. We examined VAF ranges for different groups of co-mutations, and also the possibility of discriminating them. ResultsWhile bulk analysis suggested absence of subclonal peaks and possibly neutral evolution in some cases, single-cell analysis identified co-existing subclones. The overlaps of bulk VAF ranges for co-mutations from different subclones made it difficult to separate them, even with other parameter introduced. The difference between mutation cluster and tumor subclone is accountable for the challenge in bulk clonal deconvolution, especially in case of branched evolution as shown in colorectal cancer. ConclusionComplex subclonal structures and dynamic evolution are hidden under the seemingly clonal neutral pattern at bulk level, suggesting single-cell analysis will be needed to avoid under-estimation of tumor heterogeneity. Research HighlightsO_LIBulk-level mutation abundance clusters are not equal to tumor subclones. C_LIO_LIDifferent groups of co-mutations could not be discriminated at bulk-level. C_LIO_LISingle-cell mutational analysis can identify rather than infer tumor subclones. C_LIO_LICo-existing tumor subclones may have clonal neutral appearance at bulk-level. C_LI Lay summarySystematic comparison of tumor clonal structure differences between bulk and single-cell mutational analysis is lacking. Here we performed such as study and found that complex subclonal structures and dynamic evolution are hidden under clonal neutral appearance at bulk level in liver and colorectal cancers, suggesting single-cell analysis will be needed to avoid under-estimation of tumor heterogeneity.

cancer biology↗

bioRxiv2022

An Open Microfluidic Coculture Model of Fibroblasts and Eosinophils to Investigate Mechanisms of Airway Inflammation

Interactions between fibroblasts and immune cells play an important role in tissue inflammation. Previous studies have found that eosinophils activated with interleukin-3 (IL-3) degranulate on aggregated immunoglobulin G (IgG) and release mediators that activate fibroblasts in the lung. However, these studies were done with eosinophil-conditioned media that have the capacity to investigate only one-way signaling from eosinophils to fibroblasts. Here, we demonstrate a coculture model of primary normal human lung fibroblasts (HLFs) and human blood eosinophils from patients with allergy and asthma using an open microfluidic coculture device. In our device, the two types of cells can communicate via two-way soluble factor signaling in the shared media while being physically separated by a half wall. Initially, we assessed the level of eosinophil degranulation by their release of eosinophil-derived neurotoxin (EDN). Next, we analyzed the inflammation-associated genes and soluble factors using reverse transcription quantitative polymerase chain reaction (RT-qPCR) and multiplex immunoassays, respectively. Our results suggest an induction of a proinflammatory fibroblast phenotype of HLFs following the coculture with degranulating eosinophils, validating our previous findings. Additionally, we present a new result that indicate potential impacts of activated HLFs back on eosinophils. This open microfluidic coculture platform provides unique opportunities to investigate the intercellular signaling between the two cell types and their roles in airway inflammation and remodeling.

cell biology↗

bioRxiv2022

Identification of a multipotent lung progenitor for lung regeneration

We recently showed that intravenous infusion of mouse or human, fetal or adult lung cells following conditioning of recipient mice leads to lung chimerism within alveolar and bronchiolar lineages, in distinct patches containing both epithelial and endothelial cells. We show here, using R26R-Confetti mice as donors, that these multi-lineage patches are derived from a single lung progenitor. FACS of adult mouse lung cells revealed that the putative patch-forming progenitors co-express the endothelial marker CD31 (PECAM-1) and the epithelial marker CD326 (EPCAM). Transplantation of lung cells from transgenic Cre/lox mice expressing nuclear GFP under the VEcad promoter (VEcad-Cre-nTnG), led to GFP+ patches comprising both GFP+ endothelial and epithelial cells in vivo, and in ex-vivo culture of CD326+CD31+ progenitors. Single cell RNA sequencing of CD326+CD31+ lung cells revealed a subpopulation expressing canonical epithelial and endothelial genes. Such double positive GFP+NKX2.1+SOX17+ cells were also detected by immunohistological staining in lungs of VEcad-Cre-nTnG (expressing nuclear GFP) mice in proximity to blood vessels. These findings provide new insights on lung progenitors and lung development and suggest a potential novel approach for lung regeneration. SummaryWe show in the present study, that multi-lineage regenerative patches in our transplantation model are derived from a single lung progenitor, co-expressing the endothelial marker CD31 and the epithelial marker CD326. These findings provide new insights on lung progenitors and lung development.

cell biology↗

bioRxiv2022

An improved graph representation learning method for drug-target interaction prediction over heterogeneous biological information graph

The prediction task of the relationships between drugs and targets plays a significant role in the process of new drug discovery. Computational-based strategies for predicting drug-target interactions (DTIs) are regarded as a high-efficiency way. Moreover, recent studies adopted a graph neural network (GNN) to discover underlying DTIs and achieved better performance. Although these inductive methods can straightway learn biomolecules latent representations, they have an over-smoothing phenomenon in the course of obtaining the rich neighborhood information of each node in the biological information network, which further leads to a consistent feature representation of each node. To address the above issues, a novel model, called iGRLDTI, is proposed to precisely identify new DTIs based on an improved graph representation learning strategy. Specifically, iGRLDTI first constructs a biological information graph (BIG) by calculating the biological knowledge of drugs and targets with the relationships between them. Then, an improved graph representation learning strategy is designed to capture the enriched feature representations of drugs and targets. Finally, the Gradient Boosting Decision Tree classifier is applied to predict potential DTIs. Experimental results demonstrate that iGRLDTI yields better performance by comparing it with other state-of-the-art models on the benchmark dataset. Besides, our case studies denote that iGRLDTI can successfully identify unknown DTIs according to the improved feature representations of drugs and targets.

bioinformatics↗