Search bioRxiv⌕ Search

Biology subjects

Song, T.

Publications and source records attributed to Song, T..

At least 19 recordsLinked to original sources

From Ligand Binding to Pore Opening in the Mosquito Odorant Receptor AaegOR10

Mosquito olfaction is central to host seeking and other behaviors relevant to vector ecology and disease transmission. Mosquito odorant receptors (ORs) are ligand-gated ion channels, yet how ligand binding is coupled to pore opening remains poorly understood. Here, we combine more than 400 $\mu$s of aggregate molecular dynamics simulations and adaptive sampling with Markov state models (MSMs) and electrophysiological measurements to characterize the gating dynamics of \textit{Aedes aegypti} odorant receptor 10 (AaegOR10). Interaction analysis identifies M288 as a persistent contact between $o$-cresol and S6, connecting the orthosteric pocket to a conformation-dependent hydrophobic network involving I293, F294, and F356 at the S6-S7b interface. The MSMs place the experimentally determined open and closed structures within a multistate conformational landscape and identify a principal intermediate along the dominant gating pathway. A localized S6 kink in this state is associated with weakened S6-S7b coupling and increased flexibility of the pore-lining S7b helix. Ligand binding shifts the conformational equilibrium toward open states and shortens the transition times from the closed state to both the intermediate and open states. We also identify an off-path closed state characterized by a pronounced inward displacement of S6, with phospholipid headgroup insertion observed in some sampled conformations. Mutational analysis and voltage-clamp recordings support the functional relevance of M288, the S6-S7b interface, and the kink-prone S6 region. Together, these results describe a multistate pathway linking ligand recognition to AaegOR10 pore gating.

biophysics↗

Conserved inhibitory mechanisms of root-associated bacteria towardsArabidopsis thaliana and Chlamydomonas reinhardtii

Soil bacteria colonize plants without causing visible disease, yet many suppress host growth. The prevalence and evolutionary reach of this cryptic inhibition are unknown, because phenotyping hundreds of isolates in plants is prohibitively slow. Here we use the unicellular algae Chlamydomonas reinhardtii as a scalable proxy for the green lineage. We screened 148 plant-associated bacteria on Chlamydomonas lawns. Seven of eight Chlamydomonas inhibitors also inhibited the vascular model plant Arabidopsis thaliana. In one of these dual inhibitors, Burkholderia sola MF6, transposon screening and proteomics implicated type VI secretion and tight-adherence pili, and over half the non-inhibitory mutants also failed against Arabidopsis. In the algae, infection triggers rapid deflagellation, then non-lytic regulated cell death. The algal zinc/iron transporter ZIP3 promotes inhibition, whereas the cathepsin X protease CEP12 restrains it, and ZIP3 is epistatic to CEP12. Cryptic inhibition is thus ancient and widespread, and a unicellular algae uncovers it at scale.

plant biology↗

pHaseMD4AI: Phase-Space Dynamics Dataset with Chemical and pH Perturbations for Physically and Kinetically Consistent Biomolecular AI

Protein function emerges from dynamic conformational ensembles and transitions that are challenging to characterize experimentally and computationally. Recent advances in generative AI have created new opportunities for learning molecular thermodynamics, kinetics, and conformational evolution directly from simulation data, but progress is limited by the availability of large-scale datasets that combine rigorous sampling, complete phase-space information, and diverse physicochemical perturbations. Here, we present pHaseMD4AI, a molecular dynamics dataset that combines a globally equilibrated peptide branch with a protein-scale constant-pH molecular dynamics (CpHMD) branch spanning hundreds of soluble proteins. The peptide branch includes a complete set of canonical tripeptide and tetrapeptide systems together with post-translationally modified (PTM) and protonation-state datasets, providing synchronized atomic coordinates (R), velocities (V), forces (F), and Markov state model-based kinetic annotations. An accompanying web portal (https://isb.zju.edu.cn/md4ai/) enables users to browse, visualize, and download trajectories, annotations, and metadata. As an example application, we demonstrate a sequence-based model that can predict residue-level equilibrium dihedral distributions from sequence. pHaseMD4AI provides a resource for developing and benchmarking molecular machine learning methods while supporting broader studies of biomolecular dynamics under sequence, post-translational modification, and protonation-state perturbations.

bioinformatics↗

Proteomic composition and mutual assembly of the C2a projection in vertebrate motile cilia

The central apparatus of motile cilia, consisting of central microtubules and various protein projections, is essential for dictating the ciliary movement. Although three proteins (FAP65, FAP147, and FAP70) have been localized to the C2a projection in Chlamydomonas reinhardtii, the full protein composition and functional roles of the vertebrate C2a remain inadequately defined. Here, we use three knockout mouse models corresponding to their respective homologs (Ccdc108, Mycbpap, and Cfap70) to systematically investigate their functions in vertebrates. Notably, all three knockout strains exhibit distinct phenotypes related to primary ciliary dyskinesia (PCD), including hydrocephalus and sinusitis. The ciliary incorporation of CCDC108, MYCBPAP, and CFAP70 is essential for one anothers stability, with the loss of any single component triggering C2a collapse, which destabilizes the central pair microtubules and ultimately alters the ciliary movement pattern. Furthermore, we significantly expand the vertebrate C2a proteome by identifying ARMC3 and MYCBP as additional C2a components. Collectively, our findings illuminate the proteomic composition and strict physiological requirements of the vertebrate C2a projection, providing new insights into the molecular pathogenesis of PCD. Impact StatementDeficiency in any of the interdependent C2a proteins (CCDC108, MYCBPAP, and CFAP70) collapses this central pair microtubule-associated projection, disrupting vertebrate ciliary movement and causing primary ciliary dyskinesia phenotypes in mice.

cell biology↗

Ligand Binding Kinetics, Thermodynamics, and Gating of Insect Odorant Receptor

Insect olfactory receptors (ORs) are unique ligand-gated ion channels specialized in detecting volatile molecules. The ancestral M. hrabei MhOR5 operates independently of the Orco co-receptor and exhibits broad ligand-binding capacity, yet the molecular underpinnings of its ligand specificity and gating dynamics have long remained unclear. To resolve this knowledge gap, we applied an integrated molecular dynamics enhanced sampling approach, combining collective variable machine learning, alchemical free energy calculations, and kinetic calculations, to dissect the ligand recognition and gating mechanisms of MhOR5 in complex with two key ligands: eugenol (EOL) and DEET. Our findings reveal striking differences in the binding modes of EOL and DEET, identify dual ligand binding/release pathways (solvent-facing and membrane-facing), and pinpoint two critical molecular determinants: W158 as a central gating residue regulating ligand dissociation, and a proline-disrupted S4 helix that modulates ligand specificity via water-mediated hydrogen bonding with DEET. Thermodynamic and kinetic calculations further demonstrate that DEET binds MhOR5 substantially more stably (-11.2 kcal/mol vs. -8.4 kcal/mol for EOL) and associates marginally faster, while EOL dissociates two to three orders of magnitude more rapidly. Collectively, these findings elucidate the atomic-level mechanisms of MhOR5's ligand recognition and gating, providing a foundational framework for understanding insect OR function.

biophysics↗

Transcript-Capture sequencing enriches mRNA of Mycobacterium tuberculosis from host samples

Bacterial gene expression from sites of infection are poorly studied due to low levels of bacterial mRNA present in clinical samples. Here, we develop Transcript-Capture Seq, which uses customizable biotinylated probes generated in-house to enrich bacteria-specific RNA from host samples before Next Generation Sequencing (NGS). This method results in a >200-fold increase in bacterial mRNA reads from mixed samples and allows analysis of the complete bacterial transcriptome from clinical samples. We apply Transcript-Capture to models of tuberculosis (TB) infection as well as sputum samples from TB patients. TB patients exhibit unexplained heterogeneity in disease progression, and the activity of Mycobacterium tuberculosis (Mtb) has been proposed to affect treatment response. By applying Transcript-Capture to sputum samples collected from TB patients we generate the first complete in vivo bacterial transcriptome of Mtb via NGS. Mtb from patient sputa shows upregulation of genes involved in host lipid utilization and zinc limitation, as well as a similar gene expression profile to Mtb log phase growth in vitro. Applying Transcript-Capture to clinical sputa provides a snapshot of bacterial activity directly from human patients and can be used to investigate the physiological state of bacteria surviving in vivo. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=74 SRC="FIGDIR/small/685133v1_ufig1.gif" ALT="Figure 1"> View larger version (21K): org.highwire.dtl.DTLVardef@406294org.highwire.dtl.DTLVardef@1080173org.highwire.dtl.DTLVardef@973909org.highwire.dtl.DTLVardef@21becc_HPS_FORMAT_FIGEXP M_FIG C_FIG

microbiology↗

Hugin-AstA circuitry is a novel central energy sensor that directly regulates sweet sensation in Drosophila and mouse

Taste sensation plays a crucial role in shaping feeding behavior and is intricately influenced by internal states like hunger or satiety. Despite the identification of numerous neural substrates regulating feeding behavior, the central neural substrate that linked energy-sensing and taste sensation remained elusive. Here, we identified a novel neural circuitry that could directly sense internal energy state and modulate sweet sensation in the Drosophila brain. Specifically, a subset of neuropeptidergic neurons expressing hugin directly detected elevated levels of circulating glucose via glucose transporter Glut1 and ATP-sensitive potassium channel. Upon activation, these neurons released hugin peptide and activated downstream Allatostatin A (AstA)+ neurons via its cognate receptor PK2-R1. Subsequently, the activation of AstA+ neurons then directly inhibited sweet sensation via AstA peptide and its cognate receptor AstA-R1 expressed in sweet-sensing Gr5a+ neurons. We also showed that neuromedin U (NMU), the mammalian homolog of fly hugin, served as an energy sensor to suppress sweet sensation. Therefore, these data identify hugin+ neuron as a glucose-responsive central energy-sensing module that modulates sweet sensation across species.

neuroscience↗

Engineering Versatile Two-Dimensional Nanobody-Origami Architectures for Enhanced Antiviral Activity

Considering the serious global health burden posed by pathogenic viruses, the development of effective antiviral molecules and therapeutic strategies is critical for improving human health. Designing and synthesizing a versatile and biocompatible molecular platform offering neutralization of both coronaviruses and retroviruses, two virus families that are greatly problematic to the human population, remains a significant challenge. Here, we report a programmable and versatile platform built on a two-dimensional (2D) DNA origami that enables nanoscale spatial control of multivalent nanobody (Nb) patterns for a broad-spectrum antiviral application. By site-selectively conjugating a Nb to a DNA oligonucleotide and placing multiple of them to predefined sites on 2D DNA origami, we synthesized hybrid nano-architectures with tunable Nb patterns designed to approximately match the geometric presentation of viral surface proteins. We demonstrate that such specific Nb spatial configurations significantly enhance both viral binding affinity and neutralization potency. For SARS-CoV-2, a coronavirus, a triangular Nb pattern with matched spacing with the spike proteins achieved an IC50 of 1.52 nM, representing a 171-fold improvement over monomeric Nbs. Extending this strategy to a retroviral virus, Human Immunodeficiency Virus (HIV) by utilizing a gp120 spike-binding Nb, we observed a 233-fold increase in neutralization efficiency using a patterned 2D Nb nano-architecture. These findings suggest a generalizable and versatile platform strategy for engineering potent antiviral agents through spatially optimized Nb presentation for a corresponding viral pathogen, offering a promising avenue for future antibody and Nb-based drug development.

bioengineering↗

Mitochondrial Genome Variants and Nuclear Mitochondrial DNA Segments in 7331 Individuals from NyuWa and 1KGP

Dysfunctional mitochondria are implicated in various diseases, while little is known about comprehensive characterization of mitochondrial DNA (mtDNA) in Chinese population. Here, we conducted a systematic analysis of mtDNA from 7331 samples, comprising 4129 Chinese samples (NyuWa) and 3202 samples from the 1000 Genomes Project (1KGP). We identified 7216 distinct high-quality mtDNA variants and classified them into 22 macrohaplogroups, and detected 1466 distinct nuclear mitochondrial DNA segments (NUMTs), with 88 mtDNA variants and 642 NUMTs being specific to NyuWa. The genome-wide association analyses revealed that 12 mtDNA variants were significantly correlated with 199 nuclear DNA variants. Our findings revealed that all individuals in both NyuWa and 1KGP harbored common NUMTs, and one-fifth possessed ultra-rare NUMTs, which tended to insert into nuclear gene regions. Compared to 1KGP, significant enrichment of nuclear breakpoints in long interspersed nuclear elements (LINEs) was observed for rare NUMTs in NyuWa. Overall, this study represents the first comprehensive profiling the landscape of Chinese NUMTs and offers the most extensive resource of Chinese mtDNA variants and NUMTs based on high-depth WGS to date, providing valuable reference resources for genetic research on mtDNA-related diseases.

genomics↗

Single-cell spatial map of cis-regulatory elements for disease-related genes in the macaque cortex

Single-cell spatial transcriptomes have demonstrated molecular and cellular diversity in the brain1, 2, 3, 4, 5, 6, 7, 8, but gene regulatory mechanisms underlying transcriptomic profiles and disease pathogenesis remain largely unknown in primates. Here we performed single-nucleus Assay for Transposase-Accessible Chromatin followed by sequencing (snATAC-seq) for [~]1.6 million cells from 142 cortical regions of two male cynomolgus monkeys (Macaca fascicularis), and identified distinct chromatin accessibility profiles of cis-regulatory elements (CREs) for various cell types. By integrative analysis with large-scale spatial transcriptome data, we found that these CREs showed laminar and regional preferences, with their regional accessibility exhibiting striking dependence on the regions hierarchical level. Cross-species comparison of snATAC-seq data revealed human/macaque-enriched layer-4 glutamatergic neurons and LAMP5/LHX6-expressing GABAergic neurons as well as human/macaque-biased CREs for genes related to neurodevelopment and psychiatric diseases. Importantly, risk single-nucleotide polymorphisms for many brain disorders strongly associated with human/macaque-biased CREs in glutamatergic neuronal types and those for Alzhemers disease strongly associated with CREs exclusively in microglia. Our results provided the basis for understanding the spatial gene regulatory mechanisms underlying cellular diversity and disease pathogenesis in the primate cortex.

neuroscience↗

Fast Myosin Binding Protein-C is a Vital Regulator in Young and Aged Fast Skeletal Muscle Homeostasis

BackgroundSkeletal muscle plays a vital role in voluntary motion and locomotion. Fast-twitch muscle fibers are characterized by their rapid contraction kinetics, high force generation capacity, and a distinct gene expression profile compared to slow-twitch fibers. Skeletal myosin binding protein-C (MyBP-C) paralogs, slow (sMyBP-C) and fast (fMyBP-C), interact with myosin and actin filaments within sarcomeres to modulate force development during contraction. These paralogs are differentially expressed in muscle fibers, with fMyBP-C predominantly expressed in the fast-twitch fibers. However, the role of fMyBP-C in skeletal muscle disease states and aging remains poorly understood. This study employs mouse models with fMyBP-C ablation to investigate its significance in skeletal muscle physiology. MethodsAdult skeletal muscle samples aged 2[~]7 months from male and female wild-type, db/db, MDX, ECC injury model, were used to determine the differential expression of fMyBP-C. Next, Mybpc2 knockout (C2-/-) young (3[~]5 months) and old (22 months) male mice were used to define the role of fMyBP-C in aging. Western immunoblotting was employed to analyze the expression of fMyBP-C and sMyBP-C and the phosphorylation status of sMyBP-C. The impact of C2-/- and aging on the fiber type, size, and number as well as general muscle structure was assessed by immunohistochemistry and electron microscopy. The functional effect of C2-/- and aging was measured in terms of in vivo and ex vivo muscle force generation. Lastly, RNA sequencing was performed to identify the molecular pathways dysregulated in the C2-/- mediated muscle dysfunction in young and old mice. ResultsfMyBP-C was significantly reduced with a modest compensatory upregulation of sMyBP-C in the diseased fast-twitch muscles. fMyBP-C has a significantly higher expression in the male skeletal muscles compared to females. Further studies using young male C2-/- mice showed a significant reduction in isometric tetanic force generation and relaxation rate, fiber type switching, atrophy, and altered gene expressions related to muscle function and metabolism compared to wild-type mice. Similarly, compared to their wild-type counterparts, aged male C2-/- mice display significant deficits in muscle strength, endurance, and survival rate, accompanied by changes in muscle fiber size and molecular signaling pathways critical for muscle homeostasis. ConclusionfMyBP-C is an important regulator of muscle function and homeostasis in young and aged male fast-twitch muscle fibers. The absence of fMyBP-C aggravates the effect of aging on muscle structure and function. fMyBP-C has the potential to be a therapeutic target to modulate muscle wasting caused by aging and disease.

physiology↗

Multi-omics analysis of the mechanism of alfalfa and wheat-induced rumen flatulence in Xizang sheep

This study induced a rumen flatulence model in Xizang sheep using alfalfa (HRF) and wheat grass (MRF) to explore related mechanisms. The pH of HRF and MRF groups was lower than the natural grass group. SCFA concentrations varied in different groups. Microbial analysis showed HRF and MRF groups had lower alpha and beta diversity and different compositions. Transcriptome analysis found different numbers of differentially expressed genes in each group. Spearmans correlation analysis indicated Butyrivibrio was related to GLRX and DUOX2, and Quinella to several genes.

microbiology↗

Reconstructing 3D Spatial Structures of RNA-Tomography Transcriptomes by Tensor Factorization

Cells are organized to form three-dimensional structures of complex tissues. To map the complete 3D organization of a tissue, technologies based on tissue microdissections provide deep bulk RNA sequencing of orthographically arranged cryosections of a tissue, such that the full 3D spatial structure could be inferred from deeply sequenced transcriptomes in three views projected similarly as 3D tomography. Here, we introduce CTFacTomo to learn a Collapsed Tensor Factorization for RNA tomography data from cryosections to reconstruct 3D spatially resolved gene expressions. CTFacTomo combines tensor factorization with collapsing tensor entries to match the bulk gene expressions in each cryosection, enriched by a regularization of a product graph of protein-protein interaction network and spatial graphs. In the experiments, CTFacTomo is first validated on three datasets projected from fully profiled 3D spatial gene expressions to demonstrate that CTFacTomo significantly outperforms the benchmark methods for predicting the ground-truth gene expressions based on the projected 1D spatial gene expressions of three orthographic views. CTFacTomo is then applied to two RNA tomography datasets from zebrafish embryo and mouse olfactory mucosa, respectively. In both datasets, CTFacTomo detects 3D spatial expressions of several marker genes that are consistent with the developmental or functional regions in comparison to accompanying ISH staining images. In addition, a qualitative comparison between the reconstructed zebrafish embryo gene expressions with a matched external 3D Stereo-seq dataset also suggests that CTFacTomo reconstructs more spatially coherent patterns in the whole transcriptome with state-of-the-art performance.

bioinformatics↗

S100a4+ alveolar macrophages accelerate the progression of precancerous atypical adenomatous hyperplasia by promoting fatty acid metabolism

Lung cancer is preceded by premalignant lesions, and what factors drive this transformation and the potential regulatory mode in the context of tumor initiation remain to be elucidated. In the course of precancerous lesions, we found a phasic shift in metabolic patterns. Macrophages are a heterogeneous cell population with high plasticity in the tumor microenvironment. Single-cell interaction and metabolic analyses highlighted a cellular state, S100a4+ alveolar macrophages, which exhibited distinct fatty acid metabolic activity, such as palmitic acid metabolism, at the atypical adenomatous hyperplasia stage, accompanied by an angiogenic-promoting function in a pre-neoplastic setting. These findings were reproducible in human single-cell transcriptomes and had been confirmed by histopathological staining and in vitro cell coculture assays. Taken together, the results from this study demonstrated that the S100a4+ alveolar macrophage subset contributes to tumorigenesis by altering its metabolic state, suggesting that metabolic interventions targeting this cell state in the early stage of disease may delay neoplastic transformation of the lung epithelium.

cancer biology↗

Predicting mutation-disease associations through protein interactions via deep learning

Disease is one of the primary factors affecting life activities, with complex etiologies often influenced by gene expression and mutation. Currently, wet-lab experiments have analyzed the mechanisms of mutations, but these are usually limited by the costs of wet experiments and constraints in sample types and scales. Therefore, this paper constructs a real-world mutation-induced disease dataset and proposes Capsule networks and Graph topology networks with multi-head attention (CGM) to predict the mutation-disease associations. CGM can accurately predict protein mutation-disease associations, and in order to further elucidate the pathogenicity of protein mutations, we also verified that protein mutations lead to protein structural alterations by Swiss-model, which suggests that mutation-induced conformational changes may be an important pathogenic factor. Limited by the size of the mutated protein dataset, we also performed experiments on benchmark and imbalanced datasets, where CGM mined 22 unknown protein interaction pairs from the benchmark dataset, better illustrating the potential of CGM in predicting mutation-disease associations. In summary, this paper curates a real dataset and proposes CGM to predict the protein mutations-disease associations, providing a novel tool for further understanding of biomolecular pathways and disease mechanisms.

bioinformatics↗

Occurrence, risk assessment of antibiotics and antimicrobial resistance in Escherichia coli in typical rivers of Sichuan

Worldwide interest has been generated by the presence and distribution of antibiotics and antibiotic resistance genes (ARGs) in rivers. However, there was a dearth of research on the contamination of Sichuan provinces typical rivers with antibiotics. In this study, the residual level of antibiotics in 42 national/provincial sites of 9 rivers was analyzed by UPLC-MS/MS, the ecological risk level was evaluated using risk assessment method, and the drug resistance of E. coli in water was evaluated by Kirby-Bauer method. Redundancy analysis demonstrated how residual antibiotics affect the structure of the microbial community in the Minjiang River basin (RDA). Nine rivers all contained antibiotics. Among them, the MinJiang, TuoJiang, and JiaLingJiang rivers were severely contaminated with antibiotic concentrations ranging from 0.29 to 2233.71 ng/L. The Sichuan Basins antibiotic pollution level was significantly higher when compared to other Sichuan zones, which was consistent with the regions high population density. Additionally, it was discovered that 9.77% of the E. coli isolated from 9 rivers had antibiotic resistance, and more than 5.8% of them had multidrug resistance. Moreover, Norfloxacin, amoxicillin, ampicillin, and tetracycline were the main risk factors for high ecological risk in 26 of the 42 monitoring sites. Additionally, there is a strong correlation between the microbial community change and residual antibiotic. These results offered some reference information regarding the distribution of antibiotics and ARGs in typical rivers in the Chinese province of Sichuan, and this study showed that more attention needs to be paid to antibiotic pollution in Sichuans typical rivers.

microbiology↗

Biochemical characterization of extended-spectrum β-lactamases from Akkermansia genus

Akkermansia muciniphila, a member of the Verrucomicrobiota phylum, is recognized as a key gut microbe and has emerged as a potential next-generation probiotic. Assessment of antibiotic resistance in probiotics is a prerequisite for their therapeutic application, while very few is known in Akkermansia species. Firstly, we screened eight representative class A {beta}-lactamases (36.90%-41.30% identity with known {beta}-lactamases) from the Akkermansia species, which could increase the minimum inhibitory concentration (MIC) of Escherichia coli to{beta}-lactams. Secondly, four{beta}-lactamases were purified and identified as extended-spectrum {beta}-lactamase because they exhibited hydrolase activity against 19 {beta}-lactam antibiotics from penicillin, cephalosporins, and monobactam classes. Based on sequence alignment, three-dimensional structure, and binding pocket information, we hypothesized and validated that serine at 51 position was catalytic amino acid. Thirdly, the genomic context analysis revealed the absence of mobile genetic elements or other antibiotic resistance genes surrounding {beta}-lactamase genes, suggesting that the {beta}-lactamases from Akkermansia species may not be transferable. The finding and biochemical characterization of {beta}-lactamase from Akkermansia species provide a foundational basis for the safety evaluation of Akkermansia species as probiotics.

microbiology↗

Predicting Spatially Resolved Gene Expression via Tissue Morphology using Adaptive Spatial GNNs

MotivationSpatial transcriptomics technologies, which generate a spatial map of gene activity, can deepen the understanding of tissue architecture and its molecular underpinnings in health and disease. However, the high cost makes these technologies difficult to use in practice. Histological images co-registered with targeted tissues are more affordable and routinely generated in many research and clinical studies. Hence, predicting spatial gene expression from the morphological clues embedded in tissue histological images, provides a scalable alternative approach to decoding tissue complexity. ResultsHere, we present a graph neural network based framework to predict the spatial expression of highly expressed genes from tissue histological images. Extensive experiments on two separate breast cancer data cohorts demonstrate that our method improves the prediction performance compared to the state-of-the-art, and that our model can be used to better delineate spatial domains of biological interest. Availabilityhttps://github.com/song0309/asGNN/

bioinformatics↗