Biology subjects

Tian, Y.

Publications and source records attributed to Tian, Y..

At least 91 records · Page 5Linked to original sources

bioRxiv2025

Catastrophic drought effects on water quality and phytoplankton biomass in Nanwan Reservoir (Xinyang, China)

Climate change has led to increasingly frequent and unpredictable droughts and high-temperature events, creating extreme conditions that profoundly impact the productivity of freshwater ecosystems. This study investigates the Nanwan Reservoir, a large and deep reservoir in Xinyang, China, to evaluate the effects of extreme drought events on water quality and phytoplankton production. Field investigations were conducted during both high-temperature and drought (HTD) conditions in 2019, and during non-high-temperature and drought (NHTD) conditions in 2020. Results showed that HTD conditions significantly disrupted the thermocline and oxycline structures, leading to prolonged stratification during the HTD period. Although phosphorus concentrations remained relatively stable across both periods, nitrogen levels were markedly lower during HTD events--likely due to uptake by phytoplankton--indicating a shift in nutrient limitation from phosphorus to nitrogen. Additionally, a complex relationship between environmental variables and phytoplankton biomass was observed under HTD conditions. These findings advance our understanding of primary production responses to extreme weather events in the Nanwan Reservoir and highlight the importance of incorporating such knowledge into water resource management and ecological conservation strategies.

ecology↗

bioRxiv2025

Layers of immunity: Deconstructing the Drosophila effector response

The host innate immune response relies on the cooperation of multiple defence modules. In insects and other arthropods, which have only innate immune mechanisms, four main immune-specific modules have been described in the defence against microbial invaders: the Toll pathway, the Imd pathway, the melanization response, and phagocytosis by plasmatocytes. Our present understanding of their relative importance remains fragmented as their contribution to host defence has never been simultaneously assessed across a large panel of pathogens. Here, we have taken advantage of newly-described immune mutants in a controlled genetic background to systematically delete these four immune modules individually, in pairs, or even all four simultaneously. Surprisingly, immune deficient flies simultaneously deficient in all four immune modules are viable, homozygous fertile, and display no overt morphological defects, suggesting these immune mechanisms are not strictly required for organismal development. With this new set of lines, we assessed the contribution of each module individually and collectively to host defense against five viruses, three fungi, eight Gram-positive bacteria, and eight Gram-negative bacteria. Our findings show that these four modules largely function independently and additively in host defense, although synergistic effects can occur for select pairs of modules. Our study confirmed the importance of the Imd pathway against Gram-negative bacteria and the Toll pathway against Gram-positive bacteria and fungi, largely via the induction of effectors such as antimicrobial peptides (AMPs) and Bomanins (Bom), but also reveals the important role of melanization against viruses, and the contribution of phagocytosis against various germs. Additionally, by examining microbial load kinetics in different mutants, we provide insight into how these modules contribute to tolerance or resistance against specific microbes. Our study provides insights into the architecture of the Drosophila immune system, revealing differential requirements of immune modules according to each pathogen. The set of immune deficient lines provided here offer tools to better assess the role of these immune modules in host defense.

immunology↗

bioRxiv2025

The Golgi Rim is a Precise Tetraplex of Golgin Proteins that Can Self-Assemble into Filamentous Bands

Golgin proteins have long been suspected to be organizers of the Golgi stack. Using three-dimensional super-resolution microscopy, we comprehensively localize the human golgin family at the rim of the Golgi apparatus at 10-20 nm resolution in situ. Unexpectedly, we find that the golgins are precisely organized into a tetraplex with four discrete layers, each containing a specific set of rim golgins. We observe no golgins inside the stack between its membrane-bound cisternae. Biochemically characterizing most of the golgins as isolated proteins, we find that they form anti-parallel dimers and further self-assemble into bands of multi-micron-long filaments. Based on our findings, we propose an "outside-in" physical model, the Golgin Organizer Hypothesis, in which the Golgi stack of cisternae and its overall ribbon morphology directly result from bending circumferential bands of rim golgin filaments onto a membrane surface, explaining stack formation without the need for special "stacking proteins."

cell biology↗

bioRxiv2025

Arabidopsis INHIBITOR OF GROWTH 2 (ING2) promotes flowering by regulating NuA4-dependent histone H4 acetylation levels at FT and SOC1 chromatin loci

INHIBITOR OF GROWTH (ING) proteins are chromatin readers that bind trimethylated histone H3 lysine (K) 4 (H3K4me3) marks and associate with either histone acetyltransferase or deacetylase complexes to activate or repress gene transcription, respectively. In plants, there are two types of ING proteins named ING1 and ING2. Here we report that Arabidopsis ING2 associates with multiple subunits of the histone H4 acetyltransferase complex NuA4, controls genome-wide levels of histone H4 acetylation (H4ac) and regulates different developmental processes including the initiation of flowering. Our data indicates that ING2 biological functions are largely independent of ING1 activity. We find that ING2 is recruited to the chromatin of key floral integrators such as FT and SOC1, and is required for their timely activation by modulating histone H4ac levels at these loci. Besides, our observations reveal a genetic interaction between ING2 and genes encoding relevant proteins for H3K4me3 or H2A.Z deposition, suggesting that ING2 might represent a hub for potential crosstalk between histone H4ac and these histone modifications/variants.

plant biology↗

bioRxiv2025

APC loss promotes intestinal transformation through induction of bistable stem cell states

Colorectal cancer (CRC) is a leading cause of cancer deaths, predominantly initiated by genetic inactivation of the APC tumor suppressor. Current dogma holds that APC inactivation promotes tumorigenesis via hyperactivation of the WNT pathway transcriptional effector, {beta}-CATENIN. Although {beta}-CATENIN activation is required for tumor initiation, activating mutations in {beta}-CATENIN are infrequent. Here, we ask what underlies the selective pressure for APC inactivation by comparing the oncogenic effects of APC loss to {beta}-CATENIN hyperactivation. We find that APC loss activates a {beta}-CATENIN-independent fetal intestinal transcriptional program driven by dysregulation of GSK-3 activity upon the LIM-domain protein AJUBA, a positive regulator of YAP. This results in AJUBA stabilization and downstream transcriptional activation of the YAP-driven fetal intestinal gene expression program. We find that {beta}-CATENIN- and YAP-driven transcriptional states are mutually exclusive, existing in an interchangeable bistable balance among APC-null cells. This results in more robust tumor initiation and metastatic progression downstream of APC loss relative to {beta}-CATENIN activation. Taken together, our findings explain the preferential selection for APC inactivation in CRC development and illuminate how {beta}-CATENIN- and YAP-driven gene expression programs coexist to promote tumorigenesis.

cancer biology↗

bioRxiv2025

Colorectal cancer progression to metastasis is associated with dynamic genome-wide biphasic 5-hydroxymethylcytosine accumulation.

BackgroundColorectal cancer (CRC) progression from adenoma to adenocarcinoma is associated with global reduction in 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC). DNA hypomethylation continues upon liver metastasis. Here we examine 5hmC changes upon progression to liver metastasis. Results5hmC is increased in metastatic liver tissue relative to the primary colon tumour and expression of TET2 and TET3 is negatively correlated with risk for metastasis in patients with CRC. Genes associated with increased 5-hydroxymethylcytosine show KEGG enrichment for adherens junctions, cytoskeleton and cell migration around a core cadherin (CDH2) network. Overall, the 5-hydroxymethylcyosine profile in the liver metastasis is similar to normal colon appearing to recover at many loci where it was originally present in normal colon and then spreading to adjacent sites. The underlying sequences at the recover and spread regions are enriched for SALL4, ZNF770, ZNF121 and PAX5 transcription factor binding sites. Finally, we show in a zebrafish migration assay using SW480 CRISPR-engineered TET knockout and rescue cells that reduced TET expression leads to a reduced migration frequency. ConclusionTogether these results suggest a biphasic trajectory for 5-hydroxymethyation dynamics that has bearing on potential therapeutic interventions aimed at manipulating 5-hydroxymethylcytosine levels.

cancer biology↗

bioRxiv2025

Deep generation of personalized connectomes based on individual attributes

An individuals connectome is unique. Interindividual variation in connectome architecture associates with disease status, cognition, lifestyle factors, and other personal attributes. While models to predict personal attributes from a persons connectome are abundant, the inverse task-- inferring connectome architecture from an individuals personal profile--has not been widely studied. Here, we introduce a deep model to generate a persons entire connectome exclusively based on their age, sex, body phenotypes, cognition, and lifestyle factors. Using the richly phenotyped UK Biobank connectome cohort (N=8,086), we demonstrate that our model can generate network architectures that closely recapitulate connectomes mapped empirically using diffusion MRI and tractography. We find that age, sex, and body phenotypes exert the strongest influence on the connectome generation process, with an impact approximately four times greater than that of cognition and lifestyle factors. Regional differences in the importance of measures were observed, including an increased importance of cognition in the association cortex relative to the visual system. We further show that generated connectomes can improve the training of machine learning models and reduce their predictive errors. Our work demonstrates the feasibility of inferring brain connectivity from an individuals personal data and enables future applications of connectome generation such as data augmentation and anonymous data sharing.

neuroscience↗

bioRxiv2025

Whole-brain entrainment and anticipation in fMRI motor task data.

The readiness potential (RP), also known as the Bereichtschaftspotential, is commonly observed in EEG recordings as a slow build-up of negative electrical potential prior to self-directed motor movements (Kornhuber et al. 1965). In this study, we analyzed motor task-based functional magnetic resonance (fMRI) data acquired from 262 human participants, focusing on the degree of signal phase synchronization across different frequency spans and brain regions. A new method that clusters brain regions based on instantaneous phase allowed for time resolved estimation of cross-subject phase synchronization. We show that during rest periods that precede movement, an fMRI equivalent of the RP is gradually established in a network that encompasses the bilateral supplementary motor area and parts of the cingulo-opercular network, recently described as the brains Action Mode network (Dosenbach et al. 2025). Importantly, an anticipatory negative shift in the fMRI signal was observed in both single subjects and single epochs. Tongue movement elicited strong synchronization between the orbitofrontal, ventromedial, and temporal pole cortices. Our findings suggest a link between fMRI and electrophysiological recordings of anticipatory motor events in the brain. The method presented here grades brain synchronization along an in-phase and anti-phase continuum and has applications in clinical settings, as well as for cognitive brain mapping that goes beyond anticipation. Significance statementThe readiness potential (RP) is observed in EEG recordings as a slow build-up of negative electrical potential in the supplementary motor area (SMA) starting up to two seconds before self-initiated motor movements. It is believed to play a causal role in motor preparation. Given that it is only detectable by averaging 30-40 epochs, its existence as a real signal has been challenged by new modeling approaches which suggests the early RP-component is merely a summation of stochastic noise. In this study, we demonstrate that the equivalent of the RP is present in functional magnetic resonance imaging (fMRI) data in a distributed network known as the "action mode network". The RP-equivalent is present in single subjects and single epochs.

neuroscience↗

bioRxiv2025

Comparative Analysis of Phytochemical Constituents of Ginkgo Biloba Flowers and Leaves, and Evaluation of Their Biological Activities

BackgroundsGinkgo biloba L. has attracted much attention for its unique chemical composition and pharmacological properties, and most of the research focused on Ginkgo biloba leaves (GBL). Our preliminary research found that Ginkgo biloba flowers (GBF) were superior to GBL in terms of anti-radiation activity, but the underlying cause of this discrepancy in activity remains elusive. ObjectivesThe aim of this study was to systematically compare the chemical and nutritional composition of Ginkgo biloba flowers and leaves, to further elucidate the material basis of the medicinal and nutritional value of Ginkgo biloba flowers, and to conduct further comparative studies on the anti-ferroptosis activity and anti-radiation activity of the two. MethodsIn this study, the chemical constituents of GBF and GBL were identified by UPLC-Q-orbitrap HRMS, and the contents of amino acids, fatty acids, inorganic elements, purines, and ceramides of both were determined. Meanwhile, the anti-iron death activity and anti-radiation activity of GBF and GBL were comparatively studied by cellular experiments. ResultsThe results showed that the content of chemical components such as flavonoids and other nutrients such as amino acids, fatty acids, and ceramides were richer in GBF than in GBL, and GBF was indeed superior to GBL in terms of anti-radiation and anti-ferroptosis activity. ConclusionThrough in-depth analyses of the chemical and nutritional compositions of Ginkgo biloba flowers and leaves, the present study reveals the possible material basis for the stronger activity of GBF and provides theoretical support for its application in nutraceuticals and pharmaceuticals, especially in the development of functional nutraceuticals and anti-radiation drugs with higher health benefits.

biochemistry↗

bioRxiv2025

Allosteric factors in the calcium/calmodulin-responsive kinase II hub domain determine selectivity of GHB ligands for CaMKIIa

The Ca2+/CaM-dependent protein kinase II alpha (CaMKII) is a highly important synaptic protein, which comprises a unique holoenzyme structure organized via the central hub domain. Recently, a distinct binding pocket in the CaMKII hub domain was identified for the endogenous neuromodulator {gamma}-hydroxybutyric acid (GHB) and related synthetic analogues. Key interacting residues in CaMKII were revealed, but the pronounced selectivity towards the alpha variant of CaMKII has remained unresolved. Aimed at elucidating the molecular determinants for this selectivity, we here conducted binding studies to CaMKII-HEK whole-cell homogenates using two different in-house-developed GHB-related radioligands, [3H]HOCPCA and [3H]O-5-HDC, in combination with site-directed mutagenesis. Binding to CaMKII with the smaller-type radioligand [3H]HOCPCA validated key involvement of the four known residues (His395, Arg433, Arg453 and Arg469), but also revealed a role for the upper hub flexible loop containing the CaMKII-specific residue Trp403 (Leu in all other CaMKII isozymes). Insertion of the corresponding residues (L467W/C533R) into CaMKII{beta} failed to introduce [3H]HOCPCA binding. However, with the larger-type radioligand, [3H]O-5-HDC, specific binding in CaMKII{beta} (L467W/C533R) was achieved. Thus, of the four native CaMKII isozymes, only CaMKII accommodates GHB ligands. The study identifies the CaMKII flexible pocket loop as a distantly located "allosteric" factor in determining selectivity of GHB analogues for CaMKII. It sheds light on a remarkable interplay of the entire hub cavity for accommodation of ligands, and corroborates GHB analogues as CaMKII-selective.

biochemistry↗

bioRxiv2025

ATP-sensitive isoforms of P2X5: A reevaluation of its function across species

The P2X receptor family comprises ATP-gated non-selective cation channels central to physiological processes across the nervous, immune, cardiovascular, respiratory, and reproductive systems. While P2X1, P2X2, P2X3, P2X4, and P2X7 are well-established as ATP-sensitive, P2X5 isoforms and transcript variants (TVs) have traditionally been considered ATP-insensitive, limiting their functional exploration. This study identifies previously overlooked ATP-sensitive P2X5 isoforms across diverse species. Through gene profiling and next-generation sequencing (NGS), we found that the zfP2X5TV2 isoform dominates ATP-sensitive forms in zebrafish. In mice, ATP-sensitive mP2X5G317 (mP2X5.1) comprises [~]90% of transcripts, while in rats, exon 3-containing rP2X5F191 accounts for over 70%. In human cell lines, ATP-sensitive isoforms retaining exons 3, 7, and 10 represent [~]15-30% of P2X5 transcripts. RNA-seq from human tissues confirms frequent retention of exons 3 and 7 and polymorphic exon 10 expression. ATP-sensitive P2X5 variants were also identified in chickens, bullfrogs, dogs, cows, and naked mole-rats. These findings challenge the prevailing assumption of ATP-insensitivity, highlighting the need to reassess P2X5s roles in physiological and pathological contexts.

physiology↗

bioRxiv2025

Global dissection of the impact of Alzheimer's disease on brain architecture and behavior: High resolution MRH resolves robust regional effects

Alzheimers disease (AD) affects brain regions with remarkable heterogeneity, but the precise impact of this disease on hundreds of small cortical, subcortical, and brainstem regions remains poorly defined. Here, as a prelude to testing preclinical models to prevent AD, we systematically quantified effects of human AD mutations in APP and PSEN1 on 231 regions and comprehensively evaluated changes in volume with unprecedented resolution in genetically diverse mice as a function of sex and genetic background. We studied 34 5XFAD F1 hybrids and 23 sibling controls at 14 months, evaluating learning and memory behaviors, followed by ex vivo diffusion tensor images (DTI) at 25 {micro}m resolution. We delineated 231 regions of interest (ROIs) bilaterally with high precision. Remarkably, we found bidirectional changes: marked volume increases (up to 10%) in neocortex, hippocampus, amygdala, and sensory nuclei, contrasted with decreases in striatum, pallidum, thalamus, hypothalamus, and most fiber tracts. These opposing effects are unrelated to amyloid load and are likely to reflect temporal gradients in susceptibility of ROIs. Effects are similar in both sexes but far more prominent in females. Genetic background strongly modulates penetrance of the human mutations with the AD-BXD77 F1 type having the greatest sensitivity. Light sheet microscopy and stereological analysis of NeuN+ neurons and amyloid {beta} aggregates in 22 regions revealed up to 20% loss of cells in CA3 and in anterior and intralaminar parts of the thalamus. Volumetric changes correlated with impaired fear acquisition and memory, with cases and controls often showing opposite relations between performance and regional volumes. These findings reveal unprecedented regional heterogeneity in AD progression and suggest therapeutic efficacy may vary substantially across genetic backgrounds and between sexes.

neuroscience↗

bioRxiv2025

Molecular and structural basis of a subfamily of PrfH rescuing both the damaged and intact ribosomes stalled in translation

In bacteria, spontaneous mRNAs degradation and ribotoxin-induced RNA damage are two main biological events that lead to the stall of protein translation. The ubiquitous trans-translation system as well as several alternative rescue factors (Arfs) are responsible for rescuing the stalled ribosomes caused by truncated mRNAs that lack the stop codons. To date, protein release factor homolog (PrfH) is the only factor known to rescue the stalled ribosome damaged by ribotoxins. Here we show that a subfamily of PrfH, exemplified by PrfH from Capnocytophaga gingivalis (CgPrfH), rescues both types of stalled ribosomes described above. Our in vitro biochemical assays demonstrate that CgPrfH hydrolyzes the peptides attached to P-site tRNAs when in complex with both the damaged and intact ribosomes. Two cryo-EM structures of CgPrfH in complex with the damaged and intact 70S ribosomes revealed that CgPrfH employs two different regions of the protein to recognize two different stalled ribosomes to orient the GGQ motif for peptide hydrolysis. Thus, using a combination of bioinformatic, biochemical, and structural characterization described here, we have uncovered a family of ribosome rescue factors that possesses dual activities to resolve two distinct stalled protein translation in bacteria.

biochemistry↗

bioRxiv2024

Genome-wide Mapping Eukaryotic Open Chromatin Regions with Double-Stranded DNA Deaminase

A variety of methods such as ATAC-seq and CUT&Tag have been developed for probing chromatin states. However, enzymatic cleavage of DNA in these methods hinders their multi-omics integration with other DNA profiling techniques. Here, by extending our use of dsDNA deaminase (FOOtprinting with DeamInasE, FOODIE), we introduce two new methods for mapping Chromatin Open region by DeamInasE (ChrODIE) and histone modification by ANtibody-associated DeamInasE (ANDIE), respectively. Unlike cleavage-based methods, both ChrODIE and ANDIE leave genomic DNA after deamination unfragmented and amplifiable. This offers the potential for simultaneous or downstream integration with other genomic profiling methods, such as mapping 3D genome interactions.

genomics↗

bioRxiv2024

deepBlastoid: A Deep Learning-Based High-Throughput Classifier for Human Blastoids Using Brightfield Images with Confidence Assessment

Recent advances in human blastoids have opened new avenues for modeling early human development and implantation. Human blastoids can be generated in large numbers, making them suitable for high-throughput screening, which often involves analyzing vast numbers of images. However, automated methods for evaluating and characterizing blastoid morphology are still underdeveloped. We developed a deep-learning model capable of recognizing and classifying blastoid brightfield images into five distinct quality categories. The model processes 53.2 images per second with an average accuracy of 87%, without signs of overfitting or batch eHects. By integrating a Confidence Rate (CR) metric, the accuracy was further improved to 97%, with low-CR images flagged for human review. In a comparison with human experts, the model matched their accuracy while significantly outperforming them in throughput. We demonstrate the value of the model in two real-world applications: (1) systematic assessment of the eHect of lysophosphatidic acid (LPA) concentration on blastoid formation, and (2) evaluating the impact of dimethyl sulfoxide (DMSO) on blastoids for drug screening. In the applications involving over 10,000 images, the model identified significant eHects of LPA and DMSO, which may have been overlooked in manual assessments. The deepBlastoid model is publicly available and researchers can train their own model according to their imaging conditions and blastoid culture protocol. deepBlastoid thus oHers a precise, automated approach for blastoid classification, with significant potential for advancing mechanism research, drug screening, and clinical in vitro fertilization (IVF) applications.

developmental biology↗

bioRxiv2024

ANKRD5: a key component of the axoneme required for spermmotility and male fertility

Sperm motility is essential for male fertility and depends on the structural integrity of the sperm axoneme, which features a canonical "9+2" microtubule arrangement. This structure comprises nine outer doublet microtubules (DMTs) that are associated with various macromolecular complexes. Among them, the nexin-dynein regulatory complex (N-DRC) forms crossbridges between adjacent DMTs, contributing to their stabilization and enabling flagellar bending. In this study, we investigated Ankyrin repeat domain 5 (ANKRD5, also known as ANK5 or ANKEF1), a protein highly expressed in the sperm axoneme. We found that ANKRD5 interacts with DRC5/TCTE1 and DRC4/GAS8, two key components of the N-DRC, and these interactions occur independently of calcium regulation. Male Ankrd5-/- mice exhibited impaired sperm motility and infertility. Cryo-electron tomography revealed a typical "9+2" axoneme structure with intact DMTs in Ankrd5 null sperm; however, the DMTs showed pronounced morphological variability and increased structural heterogeneity. Notably, ANKRD5 deficiency did not alter ATP levels, reactive oxygen species (ROS) levels, or mitochondrial membrane potential. These findings suggest that ANKRD5 may attenuate the N-DRCs mechanical buffering-akin to a "car bumper"-between adjacent DMTs, thereby compromising axonemal stability under high mechanical stress during vigorous flagellar beating. Graphic Abstract O_FIG O_LINKSMALLFIG WIDTH=184 HEIGHT=200 SRC="FIGDIR/small/626701v3_ufig1.gif" ALT="Figure 1"> View larger version (39K): org.highwire.dtl.DTLVardef@1ca36eforg.highwire.dtl.DTLVardef@dcafe3org.highwire.dtl.DTLVardef@9f6086org.highwire.dtl.DTLVardef@17a0a0b_HPS_FORMAT_FIGEXP M_FIG C_FIG Significance StatementMale infertility affects approximately 8%-12% of men globally, with defects in sperm motility accounting for over 80% of these cases. The axoneme, which functions as the motor apparatus of the sperm, adopts a canonical "9+2" microtubule arrangement, where the nexin-dynein regulatory complex (N-DRC) plays a critical role in providing structural support between adjacent outer microtubule doublets. Elucidating the interplay between the structural organization and protein composition of the N-DRC is essential for advancing the understanding of male reproductive biology. In this study, we identify ANKRD5 as a new component of N-DRC that is essential for maintaining normal sperm motility. These findings contribute to the molecular understanding of sperm motility and highlight ANKRD5 as a potential target for the development of novel male contraceptives.

developmental biology↗

bioRxiv2024

A Novel Efficient Algorithm for Common Variants Genotyping from Low-Coverage Sequencing Data

Low-coverage whole-genome sequencing (LC-WGS) combined with imputation represents a cost-effective genotyping strategy for genome-wide association studies (GWAS) in population genetics. In this study, the Limpute algorithm was developed specifically for genotyping from low-coverage sequencing data, it extracts variant information from low-coverage sequencing data by the novel virtual probes and subsequently performs imputation through cross-reference between samples. Compared to the currently dominant algorithm for low-coverage sequencing data, GLIMPSE2, Limpute achieved similar imputation performance within common variants (r2>0.87) while the GLIMPSE2 has a runtime approximately five times longer than that of the Limpute. Furthermore, to fully evaluate the accuracy of genotype imputation by Limpute, we utilized high-coverage whole-genome sequencing data (30x), microarray data, and high-coverage whole-exome sequencing data (30x) as validation sets respectively. The results demonstrated that Limpute has a good imputation performance for common variants using low-coverage sequencing data (1x: r2 > 0.87; 3x: r2 > 0.92; 5x: r2 > 0.93). In summary, we present a highly efficient, low-cost algorithm for genotyping from low-coverage sequencing data, offering substantial support for genetic research.

bioinformatics↗

bioRxiv2024

Deep-learning-enabled generative prevalent mutation prediction through host-to-herd in silico virus evolution

Predicting the mutation prevalence trends of emerging viruses in the real world is an efficient means to update vaccines or drugs in advance. It is crucial to develop a computational method for the prediction of real-world prevalent SARS-CoV-2 mutations considering the impact of multiple selective pressures within and between hosts. Here, a deep-learning generative framework for real-world prevalent SARS-CoV-2 mutation prediction, named ViralForesight, is developed on top of protein language models and in silico virus evolution. Through the paradigm of host-to-herd in silico virus evolution, ViralForesight reproduced previous real-world prevalent SARS-CoV-2 mutations for multiple lineages with superior performance. More importantly, ViralForesight correctly predicted the future prevalent mutations that dominate the COVID-19 pandemic in the real world more than half a year in advance with in vitro experimental validation. Overall, ViralForesight demonstrates a proactive approach to the prevention of emerging viral infections, accelerating the process of discovering future prevalent mutations with the power of generative deep learning.

evolutionary biology↗