Search bioRxiv⌕ Search

Biology subjects

Ge, R.

Publications and source records attributed to Ge, R..

At least 19 recordsLinked to original sources

RNA accessibility sequencing reveals transcriptome-wide RBP binding and RNA structure

We introduce RAD-seq (RNA Accessibility via Deamination sequencing), a simple and scalable method that quantifies RNA accessibility by protein binding in vivo. RAD-seq uses TadA8.20-mediated adenosine deamination to report the local exposure of each adenosine to its molecular environment, integrating the combined effects of base pairing, RNA folding, and RNA-binding protein (RBP) occupancy in a single experiment. RAD-seq profiles recapitulate known structural signatures, reveal high-resolution RBP footprints, report on the influence of RNA modifications, and capture dynamic remodeling of RNA accessibility during oxidative and heat-shock stress. By integrating RAD-seq with RNA Bind-n-Seq motifs, we detected local enrichment of the RNA N6-methyladenosine (m6A) methyltransferase recruiter TRA2A following knockdown of the exon junction anchor protein EIF4A3, revealing a potential mechanism that links RNA accessibility to enhanced TRA2A recruitment and m6A deposition. Extending this framework, we introduce an unsupervised statistical model that combines RAD-seq with motif priors to infer binding landscapes for multiple RBPs, reproducing the information from numerous CLIP-seq datasets from a single RAD-seq experiment. These results indicate RAD-seq as a generalizable approach for joint profiling of RNA structure and protein occupancy, defining enzymatic RNA accessibility as a new dimension to investigate regulatory elements on RNA.

biochemistry↗

Medication and Developmental Stage Shape the Brain-Age Gap in ADHD

Identifying robust biomarkers for Attention-Deficit/Hyperactivity Disorder (ADHD) remains challenging. The brain-age gap (BAG), which is the difference between predicted biological brain age and chronological age, has been proposed as a marker of atypical neurodevelopment, yet findings have been inconsistent across studies, limiting its clinical translation. We hypothesized that two major sources of heterogeneity, medication status and developmental stage, contribute to these discrepancies. To test this, we analyzed structural MRI data from a large multi-site cohort (N = 947; ages 7-26 years). Individuals with ADHD were stratified by medication history (medicated vs. unmedicated) and developmental stage (childhood vs. adolescence) and compared with matched typically developing controls. We estimated global- and network-level BAG using publicly available pretrained models. Our results revealed that BAG differences were contingent on medication status, with abnormalities predominantly observed in unmedicated individuals. Developmental stage further moderated the presence of this effect: BAG abnormalities were present significantly in children with ADHD but not in adolescents. Network-level analyses demonstrated spatial specificity, with significant effects in visual, ventral attention, and dorsal attention networks. Together, these findings provide compelling evidence that BAG in ADHD is both medication- and development-dependent, highlighting the need to account for treatment exposure and developmental stage when evaluating BAG as a biomarker and further underscoring the importance of early intervention during childhood.

neuroscience↗

Selective PPAR-α activation with pemafibrate attenuates macrophage-mediated progression of calcific aortic valve disease

BACKGROUNDCalcific aortic valve disease (CAVD) compromises valve compliance and cardiac hemodynamics leading to aortic stenosis (AS) and cardiovascular dysfunction. With treatment for severe AS limited to valve replacement and no effective pharmacotherapies, new interventions are urgently needed for patients. This study evaluated pemafibrate, a selective peroxisome proliferator-activated receptor alpha (PPAR) activator as a novel therapeutic for CAVD and AS. METHODS AND RESULTSIn an aortic valve wire injury (AVWI) model of AS in Ldlr-/- mice, pemafibrate administration (0.2 mg/kg/day) for 15 weeks improved aortic valve function and reduced valvular calcification by 39% (p<0.001), accompanied by reduced leaflet inflammation and CD68 macrophage infiltration. These effects were independent of changes in plasma triglyceride levels. In vitro, pemafibrate suppressed inflammation-mediated calcification of primary human valvular interstitial cells (VICs) by modulating macrophage-derived secreted factors, identifying macrophage-VIC crosstalk as a key disease mechanism. Direct treatment of macrophages with pemafibrate, or exposure to serum from pemafibrate-treated participants in the PROMINENT randomized controlled trial, shifted macrophages toward a less inflammatory and less chemotactic phenotype. Proteomic analyses of patient serum substantiated these findings by reflecting a systemic reduction in inflammatory parameters and monocyte activation. Network integration of the in vitro derived pemafibrate-responsive proteome with human calcified AV tissue proteomes identified aberrant protein translation (GNB2L1, GSPT1) and disrupted bioenergetics (MYDGF, PDIA4) as potential clinically relevant pemafibrate-responsive pathways and effector proteins relevant to AS progression. CONCLUSIONSPemafibrate slows experimental AS progression and valve calcification through modulation of macrophage-VIC crosstalk, independent of lipid lowering. These findings support further evaluation of pemafibrate as a potential pharmacological approach for CAVD and support further testing in randomized clinical trials.

pharmacology and toxicology↗

Single-nucleotide resolution mapping of m6A of zebrafish mRNAs in early embryonic development links m6A modifications to the maternal-to-zygotic transition

N6-methyladenosine (m6A) is the most abundant epigenetic modification on eukaryotic mRNA and has important roles in gene expression. While m6A distribution and localization have been defined in zebrafish using antibody-based methods, no research has been published on the distribution of m6A during zebrafish embryogenesis with single-base precision. Here, we present the transcriptome-wide m6A landscape at single-nucleotide resolution during zebrafish early embryonic development using m6A-selective allyl chemical labeling and sequencing (m6A-SAC-seq). We identify over 36,000 high-confidence unique m6A sites that are primarily located in the DRACH consensus motif in long internal exons, 3UTRs, and near stop codons. Our results show a >4-fold decrease in m6A levels from 256-cell to 50% epiboly, clearly linking this methylation to zebrafish maternal-to-zygotic transition. We highlight that differential methylation between developmental stages was observed in transcripts associated with mRNA metabolism, translational processing, and organ development.

genomics↗

N6-methyladenosine regulation of mRNA translation is essential for early human erythropoiesis

N6-methyladenosine (m6A) is an abundant modification of mRNA with important regulatory roles in normal and malignant hematopoiesis. We previously reported that in human erythroid leukemia (HEL) cells, m6A mRNA marking selectively regulates translation of essential erythropoiesis genes required for in vitro differentiation and human erythroid colony formation. Here, we further investigated the timing and nature of requirement for m6A-methyltransferase (MTase) activity during human erythropoiesis, using a standardized in vitro erythroid differentiation assay for hHSPCs. We identified two critical m6A regulated developmental windows in BFU-E and during the transition from CFU-E to proerythroblasts. These windows of m6A-MTase requirement coincide with rising global m6A levels, which peak in proerythroblasts. After proerythroblast formation, however, m6A -MTase activity is dispensable for differentiation, proliferation, and survival. In BFU-E, m6A-MTase promotes proliferation but is dispensable for differentiation, while, in CFU-E, both m6A -MTase and the YTHDF family of m6A readers are essential for differentiation to proerythroblasts. Mechanistically, in CFU-E, m6A MTase activity enhances translation of ribosomal and oxidative phosphorylation (OXPHOS) genes, thereby elevating global protein synthesis rates and enabling efficient erythroblast formation. We propose that this form of translational regulation by m6A emerged as an evolutionary adaptation to meet the high translational demands of human erythropoiesis.

developmental biology↗

The Cortical Temporal Axis: MEG-Based Cross-Frequency Gradients with Biological Anchors

Neural oscillations have long been used to characterize the temporal dynamics of individual brain regions, yet a parsimonious, system-level representation that integrates multi-rhythmic activity has been lacking. Using source-localized resting-state magnetoencephalography (MEG) data, we computed regional power spectra and assessed spectral similarity, then applied diffusion map embedding to construct cross-frequency neurophysiological gradients that place whole-brain oscillations within a unified, low-dimensional coordinate system. We found the first three gradients accounted for over 40% of the variance, remained stable across individuals, and aligned with established functional, structural, and geometric cortical axes. Computational modeling showed that these gradients reflect local excitation-inhibition balance, while multimodal analyses revealed strong associations with neurotransmitter receptor distributions, cytoarchitecture, and cell-type-specific gene expression. Lifespan analyses further demonstrated systematic gradient reorganization, with distinct cognitive mappings onto functions such as language, memory, and multisensory integration. Clinically, Parkinsons disease patients displayed disrupted gradients, particularly in regions linked to language and social cognition. Finally, these gradient patterns exhibited high test-retest reliability. These findings establish MEG-derived neurophysiological gradients as robust, low-dimensional representations of cortical organization, offering a biologically grounded framework for studying brain aging and disease.

neuroscience↗

End-to-end high-throughput single-cell proteomics via SPRINT and dual-spray LC-MS

Single-cell proteomics (SCP) enables direct measurement of protein heterogeneity but remains constrained by throughput and limited applicability to primary tissues. Here, we present an integrated workflow developed to address both challenges. We engineered SPRINT, an AI-powered bioprinting platform that prepares more than 10,000 single cells per day, over tenfold faster than existing systems, while maintaining stability and enabling identification of over 6,000 proteins from individual HeLa cells. To expand analytical capacity, we further designed a dual-spray tandem direct injection (TDI) LC system that parallelizes non-analytical steps with peptide separation, doubling MS utilization efficiency and enabling 168 label-free SCP runs per day without sensitivity loss. Together with optimized tissue-dissociation protocols, this integrated workflow enabled creation of the first mouse tissue-derived SCP atlas, profiling >1,000 single cells across six organs. These advances establish SCP as a scalable platform ready for broad biological and biomedical applications.

systems biology↗

Catch and release of sialoglycoRNAs enables sequencing-based profiling across cellular and extracellular material

Glycosylated RNAs (glycoRNAs) represent a recently discovered class of small RNAs, but their systematic characterization has been limited by reliance on metabolic chemical reporters and high RNA input requirements. Here we present rPAL sequencing (rPAL-seq), a sensitive and selective platform for de novo discovery of sialoglycoRNAs. rPAL-seq combines enhanced periodate oxidation of sialic acids with a capture-release workflow and optimized library construction using poly(A) extension coupled with template-switching reverse transcription. The method enabled reproducible profiling from less than 100 ng of input RNA, corresponding to less than 2% of the material required by previous approaches. When applied across 13 human cell lines, rPAL-seq identified lineage-associated glycoRNA patterns alongside a conserved core dominated by uridine-rich snRNAs and snoRNAs, with modification signatures implicating glycosylation on acp3U or related uridine-based modifications. Extending to extracellular vesicles and non-vesicular nanoparticles, rPAL-seq revealed secreted glycoRNA profiles distinct from those of the cellular fraction. rPAL-seq provides a robust, scalable strategy for glycoRNA profiling, opening new avenues for studying this emerging biopolymer.

molecular biology↗

Stem-like Prostate Remodeling in Obesity Mediates Resistance to 5α-Reductase Inhibition Therapy in BPH

Benign prostatic hyperplasia (BPH) is closely associated with obesity and metabolic syndrome. Although 5-reductase inhibitors (5ARIs) are widely used to treat BPH, their effectiveness varies significantly, particularly in obese patients. However, the mechanisms by which obesity affects prostate growth and modulates therapeutic responses remain poorly understood. In mice, highfat diet (HFD) increased prostate weight and induced SRD5A2-independent epithelial remodeling with expansion of proximal urethral luminal epithelial cells. ScRNA-seq of prostate tissues from Srd5a2null and control mice on HFD showed enrichment of proximal stem-like populations, reduced androgen signaling, and activation of WNT and NOTCH pathways via heightened stromal-epithelial crosstalk. Clinically, BMI gain, regardless of 5ARI use, correlated with transition zone growth, stronger activity of stem-cell signatures, reduced SRD5A2, and androgen signaling downregulation. Xenium spatial profiling indicated that BMI gain expands stem-like epithelial populations and a NOTCHenriched periepithelial stromal niche surrounding epithelial compartments, supporting stemness. Patients with substantial BMI gain were less responsive to 5ARI, whereas weight control plus 5ARI therapy synergistically improved outcomes. In conclusion, BMI gain promotes proximal prostate enlargement through an SRD5A2-independent stem-like cell mediated pathway that blunts 5ARI efficacy. Combining pharmacotherapy with weight control yields superior efficacy and supports individualized management of BPH.

pathology↗

Neurodevelopmental Impact of Bipolar Disorder Genetic Risk on Cortical Thickness and Network Topology in Adolescents

Bipolar disorder (BD) is a highly heritable psychiatric condition characterized by recurrent mood episodes that commonly manifest during adolescence. Although polygenic risk scores (PRS) effectively quantify genetic susceptibility to BD, the neurobiological correlates of this genetic risk during adolescent brain development remain unclear. Using normative modeling and longitudinal neuroimaging data from the Adolescent Brain Cognitive Development (ABCD) cohort (N = 4519), we examined cortical thickness (CT) deviations and structural covariance network (SCN) alterations in adolescents stratified into high and low BD genetic risk groups based on PRS. Adolescents with high PRS for bipolar disorder exhibited greater cortical thickening in the inferior frontal gyrus and primary visual cortex, whereas those with low PRS demonstrated greater thickening in the posterior cingulate cortex and middle precentral gyrus. These patterns suggest PRS-related variations in cortical maturation, potentially reflecting distinct neurodevelopmental trajectories associated with genetic susceptibility to bipolar disorder. Furthermore, high PRS individuals displayed altered SCN topology, characterized by decreased local clustering and enhanced global network efficiency. Longitudinal data show that these abnormal regions exhibit atypical developmental trajectories, accompanied by a global reorganization of network topology. Additionally, high genetic risk was associated with lifestyle factors, especially correlated with increased positive expectancies toward substance use. Eventually, we found three potential BD risk genes during adolescence, including PLEKHA2, ZSCAN31 and ANK3, encoding the function of the postsynaptic membrane and synaptic membrane. These findings elucidate early neurodevelopmental deviations linked to genetic risk for BD, highlighting potential biomarkers for early identification and targeted interventions.

neuroscience↗

Adolescents with major depression featured by sensory-association subtyping show divergent information dynamics and streams

Adolescent major depressive disorder (MDD) exhibits complex and heterogeneous alterations of brain functional organization. To understand the neurobiological basis of adolescent MDD, we adopted resting-state functional MRI data and used various matrix decomposition approaches to obtain the organization gradients, temporal dynamics, and information streams. With clustering sensory-association gradient features in our exploratory sample (NMDD = 250 and NControls = 203), we identified two MDD subtypes. Subtype 1 was characterized by sensory contraction and subtype 2 was associated with association expansion. In addition, two subtypes showed divergent bottom-up and top-down information flows in sensory and association areas using temporal dynamics analysis. These subtypes exhibit distinct age-related changes and reorganization trajectories along sensory-association and auditory-visual axes, highlighting that cortical information flow patterns systematically vary and relate differently to sensory integration, cognitive complexity, and aging. These network distinctions are linked to clinical severity and molecular mechanisms. Subtype 1 is predominantly associated with early neurodevelopmental abnormalities and emotional regulation deficits, while Subtype 2 is more related to synaptic dysfunction and reduced neuronal excitability. These results could be largely replicated in another independent sample (NMDD = 73 and NControls = 28). We therefore construct a sensory-association dual functional framework to characterize MDD heterogeneity in adolescent MDD. Itl integrates cortical hierarchy, developmental trajectories, and genetic influences, offering novel insights into MDD pathophysiology and providing a theoretical foundation for precision psychiatry, facilitating personalized diagnosis and intervention strategies.

neuroscience↗

Generalizability of Normative Models of Brain Morphometry Across Distinct Ethnoracial Groups

Normative modeling of brain morphometric data can inform about the clinical significance of deviations from typical patterns in brain structure. Their usefulness, however, is dependent on their applicability to diverse ethnoracial groups. With this in mind, we developed age- and sex-specific normative models for cortical thickness, surface area, and subcortical volumes using brain scans from 37,407 healthy individuals from a diverse international sample. Here we demonstrate the validity of these models in diverse and distinct populations. Specifically, we tested these pre-trained models on independent samples of healthy individuals that either self-identified as Black, South Asian, East Asian Chinese, East Asian Japanese, or we categorized as African, Admixed American, East Asian, and European based on their genetic ancestry. Regardless of ethnoracial definition, the performance of the pretrained models in these samples was exceptionally high; the relative mean absolute error for each regional brain morphometry measure was less than 10% across all the distinct ethnoracial groups. These findings affirm the broad applicability of our models, ensuring that brain morphometry assessments using these models are accurate and reliable for individuals regardless of background. This broad applicability has significant implications for advancing personalized medicine and improving health outcomes in diverse populations.

bioinformatics↗

Brain-Age Prediction: Systematic Evaluation of Site Effects, and Sample Age Range and Size

Structural neuroimaging data have been used to compute an estimate of the biological age of the brain (brain-age) which has been associated with other biologically and behaviorally meaningful measures of brain development and aging. The ongoing research interest in brain-age has highlighted the need for robust and publicly available brain-age models pre-trained on data from large samples of healthy individuals. To address this need we have previously released a developmental brain-age model. Here we expand this work to develop, empirically validate, and disseminate a pre-trained brain-age model to cover most of the human lifespan. To achieve this, we selected the best-performing model after systematically examining the impact of site harmonization, age range, and sample size on brain-age prediction in a discovery sample of brain morphometric measures from 35,683 healthy individuals (age range: 5-90 years; 53.59% female). The pre-trained models were tested for cross-dataset generalizability in an independent sample comprising 2,101 healthy individuals (age range: 8-80 years; 55.35% female) and for longitudinal consistency in a further sample comprising 377 healthy individuals (age range: 9-25 years; 49.87% female). This empirical examination yielded the following findings: (1) the accuracy of age prediction from morphometry data was higher when no site harmonization was applied; (2) dividing the discovery sample into two age-bins (5-40 years and 40-90 years) provided a better balance between model accuracy and explained age variance than other alternatives; (3) model accuracy for brain-age prediction plateaued at a sample size exceeding 1,600 participants. These findings have been incorporated into CentileBrain [https://centilebrain.org/#/brainAGE2], an open-science, web-based platform for individualized neuroimaging metrics.

bioinformatics↗

Normative Modeling of Brain Morphometry Across the Lifespan using CentileBrain: Algorithm Benchmarking and Model Optimization

We present an empirically benchmarked framework for sex-specific normative modeling of brain morphometry that can inform about the biological and behavioral significance of deviations from typical age-related neuroanatomical changes and support future study designs. This framework was developed using regional morphometric data from 37,407 healthy individuals (53% female; aged 3-90 years) following a comparative evaluation of eight algorithms and multiple covariate combinations pertaining to image acquisition and quality, parcellation software versions, global neuroimaging measures, and longitudinal stability. The Multivariate Factorial Polynomial Regression (MFPR) emerged as the preferred algorithm optimized using nonlinear polynomials for age and linear effects of global measures as covariates. The MFPR models showed excellent accuracy across the lifespan and within distinct age-bins, and longitudinal stability over a 2-year period. The performance of all MFPR models plateaued at sample sizes exceeding 3,000 study participants. The model and scripts described here are freely available through CentileBrain (https://centilebrain.org/).

bioinformatics↗

Normative modeling of brain morphometry in Clinical High-Risk for Psychosis

ImportanceThe lack of robust neuroanatomical markers of psychosis risk has been traditionally attributed to heterogeneity. A complementary hypothesis is that variation in neuroanatomical measures in the majority of individuals at psychosis risk may be nested within the range observed in healthy individuals. ObjectiveTo quantify deviations from the normative range of neuroanatomical variation in individuals at clinical high-risk for psychosis (CHR-P) and evaluate their overlap with healthy variation and their association with positive symptoms, cognition, and conversion to a psychotic disorder. Design, Setting, and ParticipantsClinical, IQ and FreeSurfer-derived regional measures of cortical thickness (CT), cortical surface area (SA), and subcortical volume (SV) from 1,340 CHR-P individuals [47.09% female; mean age: 20.75 (4.74) years] and 1,237 healthy individuals [44.70% female; mean age: 22.32 (4.95) years] from 29 international sites participating in the ENIGMA Clinical High Risk for Psychosis Working Group. Main Outcomes and MeasuresFor each regional morphometric measure, z-scores were computed that index the degree of deviation from the normative means of that measure in a healthy reference population (N=37,407). Average deviation scores (ADS) for CT, SA, SV, and globally across all measures (G) were generated by averaging the respective regional z-scores. Regression analyses were used to quantify the association of deviation scores with clinical severity and cognition and two-proportion z-tests to identify case-control differences in the proportion of individuals with infranormal (z<-1.96) or supranormal (z>1.96) scores. ResultsCHR-P and healthy individuals overlapped in the distributions of the observed values, regional z-scores, and all ADS vales. The proportion of CHR-P individuals with infranormal or supranormal values in any metric was low (<12%) and similar to that of healthy individuals. CHR-P individuals who converted to psychosis compared to those who did not convert had a higher percentage of infranormal values in temporal regions (5-7% vs 0.9-1.4%). In the CHR-P group, only the ADSSA showed significant but weak associations (|{beta}|<0.09; PFDR<0.05) with positive symptoms and IQ. Conclusions and RelevanceThe study findings challenge the usefulness of macroscale neuromorphometric measures as diagnostic biomarkers of psychosis risk and suggest that such measures do not provide an adequate explanation for psychosis risk. Key pointsO_ST_ABSQuestionC_ST_ABSIs the risk of psychosis associated with brain morphometric changes that deviate significantly from healthy variation? FindingsIn this study of 1340 individuals high-risk for psychosis (CHR-P) and 1237 healthy participants, individual-level variation in macroscale neuromorphometric measures of the CHR-P group was largely nested within healthy variation and was not associated with the severity of positive psychotic symptoms or conversion to a psychotic disorder. MeaningThe findings suggest the macroscale neuromorphometric measures have limited utility as diagnostic biomarkers of psychosis risk.

neuroscience↗

Neuroimaging profiling identifies distinct brain maturational subtypes of youth with mood and anxiety disorders

Mood and anxiety disorders typically begin in adolescence and have overlapping clinical features but marked inter-individual variation in clinical presentation. The use of multimodal neuroimaging data may offer novel insights into the underlying brain mechanisms. We applied Heterogeneity Through Discriminative Analysis (HYDRA) to measures of regional brain morphometry, neurite density, and intracortical myelination to identify subtypes of youth, aged 9-10 years, with mood and anxiety disorders (N=1931) compared to typically developing youth (N=2823). We identified three subtypes that were robust to permutation testing and sample composition. Subtype 1 evidenced a pattern of imbalanced cortical-subcortical maturation compared to the typically developing group, with subcortical regions lagging behind prefrontal cortical thinning and myelination and greater cortical surface expansion globally. Subtype 2 displayed a pattern of delayed cortical and subcortical maturation indicated by higher cortical thickness and subcortical volume and lower cortical surface area expansion and myelination compared to the typically developing group. Subtype 3 showed evidence of atypical brain maturation involving globally lower cortical thickness and surface coupled with higher myelination and neural density. Subtype 1 had superior cognitive function in contrast to the other two subtypes that underperformed compared to the typically developing group. Higher levels of parental psychopathology, family conflict, and social adversity were common to all subtypes, with subtype 3 having the highest burden of adverse exposures. These analyses comprehensively characterize pre-adolescent mood and anxiety disorders, the biopsychosocial context in which they arise, and lay the foundation for the examination of the longitudinal evolution of the subtypes identified as the study sample transitions through adolescence.

pathology↗

S373P Mutation of Spike Protein in SARS-CoV-2 Stabilizes Omicron

A cluster of several newly occurring mutations on Omicron are found at the {beta}-core region of spike proteins receptor-binding domain (RBD), where mutation rarely happened before. Notably, the binding of SARS-CoV-2 to human receptor ACE2 via RBD happens in a dynamic airway environment, where mechanical force caused by coughing or sneezing occurs and applies to the proteins. Thus, we used atomic force microscopy-based single-molecule force spectroscopy (AFM-SMFS) to measure the stability of RBDs and found that the mechanical stability of Omicron RBD increased by ~20% compared with the wild-type. Molecular dynamics simulations revealed that Omicron RBD showed more hydrogen bonds in the {beta}-core region due to the closing of the -helical motif caused primarily by mutation S373P, which was further confirmed experimentally. Moreover, the binding ability of Omicron to ACE2 is promoted with a stabilized RBD. This work reveals the effect of the highly conserved mutation S373P which is present in most Omicron subvariants, including BA.1-5, BQ. 1, XBB, and CH.1.1.

biophysics↗

HybAVPnet: a novel hybrid network architecture for antiviral peptides identification

MotivationThe virus poses a great threat to human production and life, thus the research and development of antiviral drugs is urgently needed. Antiviral peptides play an important role in drug design and development. Compared with the time-consuming and laborious wet chemical experiment methods, accurate and rapid identification of antiviral peptides using computational methods is critical. However, it is still challenging to extract effective feature representations from the sequences for the accurate identification of antiviral peptides. ResultsThis study introduces a novel two-step approach, named HybAVPnet, with a hybrid network architecture to identify antiviral peptides based on neural networks and traditional machine learning methods. Firstly, eighteen kinds of features are extracted to predict labels and probabilities by the neural network classifier and LightGBM classifier, respectively. Secondly, the support vector machine classifier is carried out using the predicted probability of the first step to make the final prediction. The experimental result shows HybAVPnet can achieve better and more robust performance compared with the state-of-the-art methods, especially on independent datasets, which makes it useful for the research and development of antiviral drugs. Meanwhile, it can also be extended to other peptide recognition problems because of its generalization ability. Availability and implementationThe predicted model could be downloaded from: https://github.com/greyspring/HybAVPnet Contactgespring@hdu.edu.cn; yp.cai@siat.ac.cn Supplementary informationSupplementary data are available at Bioinformatics online.

bioinformatics↗