Biology subjects

Wang, Q.

Publications and source records attributed to Wang, Q..

At least 73 records · Page 4Linked to original sources

bioRxiv2026

Convergent genome- and gene-level constraints shape repeated environmental adaptation in grasses

Grasses (Poaceae) dominate terrestrial ecosystems and sustain global food security, yet the genomic principles enabling their repeated adaptation to extreme environments remain unresolved. Combining dense phylogenomic sampling, global environmental data, and genomic large language models (gLLMs), we characterize the mutational targets underlying environmental adaptation across 707 genomes from 569 species spanning 17 climate zones. We identify 19-30 phylogenetically independent transitions into extreme temperature, water, and soil environments, accompanied by convergent shifts in genome-scale molecular properties, including the Nitrogen-to-Carbon balance and the biosynthetic cost of the proteome. Our gLLMs-informed phylogenetic mixed modeling framework identifies 330 genes that repeatedly underlie distinct axes of environmental adaptation, highlighting the importance of protein modification and localization within extracellular and organellar compartments. Overlaying independent convergent adaptation tests identifies 17 high-confidence candidates for further characterization. Together, our results show that grass adaptation is canalized by layered constraints at genome-wide and gene-specific scales, producing predictable evolutionary trajectories.

evolutionary biology↗

bioRxiv2026

PepForge: Hierarchical HELM-Based Peptide Generation

Peptides carrying special connections such as macrocyclizations and various other structural modifications constitute a major class among peptide therapeutics, yet their chemical space remains largely inaccessible to computational generation methods. Here we present PepForge, a deep learning platform for peptide generation that exploits Hierarchical Editing Language for Macromolecules (HELM) notation to access the chemical space of modified peptides, through a Layout-Content-Connection (LCC) cascade decomposing the generation task into block layout, monomer content, and special connection prediction. The LCC cascade is trained on 383,817 HELM peptides covering 425 monomers and nine connection types. Beyond de novo generation, the LCC cascade supports masked infilling for targeted scaffold modification and multi-level constrained generation. Both the monomer library and the connection-type set support user-defined extensions for exploring a broader chemical space. The prediction module is decoupled from generation and accepts arbitrary scoring heads for downstream tasks. As a demonstration, we built an antimicrobial potency ensemble predictor trained on 11,026 peptides with minimum inhibitory concentration (MIC) values, alongside the external PeptiVerse predictor. Applied at scale, we generated 4.78 million novel HELM peptides and obtained 799 structurally novel hit antimicrobial peptide (AMP) candidates after potency and safety filtering. All code, pre-trained models, and a web interface for interactive use are publicly available at https://github.com/wqx1999/PepForge.

bioinformatics↗

bioRxiv2026

Androgen depletion increases sensitivity to effort-related costs and alters mesoaccumbal circuit function in male mice

BackgroundAndrogen deficiency in males is associated with reduced motivation, fatigue, and decreased goal-directed behavior, yet the neural mechanisms underlying these changes remain poorly understood. Dopamine signaling within the nucleus accumbens (NAc) plays a central role in regulating effort-based decision making. Here, we tested the hypothesis that loss of testicular hormones alters mesoaccumbal dopamine function to increase sensitivity to effort-related costs. MethodsMale mice underwent orchiectomy (ORX) either before puberty onset or in adulthood. Effort-based decision making was assessed using a progressive ratio 1 closed economy (PR1-CE) task. Dopamine-related function was assessed using systemic haloperidol administration and high-performance liquid chromatography to measure dopamine and metabolites, while whole-cell recordings were used to assess intrinsic excitability of NAc spiny projection neurons (SPNs). ResultsORX increased sensitivity to effort costs, reflected by a shift toward energy-efficient responding while maintaining overall food intake. These behavioral changes were accompanied by reduced responsiveness to haloperidol. Postpubertal ORX increased dopamine content and reduced metabolite-to-dopamine ratios in the NAc, consistent with reduced dopamine turnover, whereas prepubertal ORX did not affect dopamine measures. Prepubertal ORX selectively reduced excitability of NAc core D1R+ SPNs, while postpubertal ORX increased excitability across both D1R+ and D1R- populations. ConclusionsAndrogen depletion increases effort cost sensitivity and is associated with alterations in mesoaccumbal circuit function. Although behavioral effects were similar following pre- or postpubertal ORX, distinct neurochemical and cellular adaptations were observed, suggesting developmental timing influences neural adaptations to androgen depletion. These findings provide insight into neural mechanisms linking androgen deficiency to motivational deficits.

neuroscience↗

bioRxiv2026

A Monoclonal Antibody Panel to Track Ongoing Antigenic Evolution of SARS-CoV-2 Variants

SARS-CoV-2 continues to evolve, generating successive variants with enhanced resistance to antibody-mediated neutralization. However, many existing monoclonal antibody (mAb) repertoires were derived from exposures to ancestral or early Omicron strains and thus incompletely capture the antigenic landscape of newly emerging strains. Here, we generated a mAb panel by immunizing naive, unimprinted VelocImmune humanized mice with spike from the KP.3.1.1 variant, and isolated and characterized 11 neutralizing antibodies targeting diverse epitopes of recently evolved antigenic variants. We then integrated these antibodies with previously characterized mAbs to establish an expanded toolbox for profiling antigenic evolution among emerging variants. Application of this panel to currently emerging SARS-CoV-2 variants provided mechanistic insight into recent antigenic evolution, highlighting the potential role of altered receptor-binding domain conformation. Our data also revealed clear differences in antibody-evasion profiles between JN.1-derived lineages and BA.3.2 lineages. Together, our findings establish an updated mAb panel for monitoring SARS-CoV-2 antigenic evolution.

microbiology↗

bioRxiv2026

S-nitrosylation of protein kinase A is required for its activation by GPCRs

Stimulation of many G protein-coupled receptors (GPCRs) increases cyclic adenosine monophosphate (cAMP) and nitric oxide (NO). While cAMP-dependent activation of protein kinase A (PKA) is a central regulatory mechanism, a parallel role for NO in GPCR transduction has not been established. Here we show that upon stimulation of multiple GPCRs in heart, brain, and fat, the regulatory subunits of PKA undergo enzymatic S-nitrosylation by SNO-CoA-assisted nitrosylase (SCAN). S-nitrosylation by SCAN is required for dissociation of PKA holoenzymes and activation of PKA. In transgenic and cardiovascular disease models, impaired adrenergic stimulation is identified with deficient S-nitrosylation of PKA, which can be rescued by the drug sodium nitroprusside. Our work suggests a new understanding of GPCR physiology with direct application to the clinical setting.

physiology↗

bioRxiv2026

Chronic Mild Stress Impairs Hippocampal Myelination through SOX6-Dependent Dysfunction of Oligodendrocyte Lineage Cells

Chronic stress induces structural and functional changes in the brain, increasing susceptibility to major depressive disorder and other mental illnesses. Myelination deficits are a key pathological feature of stress-related disorders, yet the molecular mechanisms linking chronic stress to oligodendrocyte dysfunction remain poorly understood. Here, we used single-nucleus multiome sequencing to map gene expression and chromatin-accessibility remodeling in the hippocampus of mice exposed to chronic unpredictable mild stress (CUMS), a model that simulates key features of human daily stressors. CUMS induced broad molecular reprogramming across hippocampal cell populations, with stress-responsive gene networks significantly enriched for depression-associated genes. The oligodendrocyte lineage showed heightened vulnerability, with CUMS preferentially disrupting immature OPC/intermediate states and impairing OPC migration, OPC-to-ODC lineage progression, intercellular communication, and myelination. Integrated multiomic analysis identified stage-specific cis-regulatory elements and stress-sensitive gene regulatory networks, converging on SOXD transcription factors, particularly SOX5 and SOX6, as key regulators of OPC dysfunction. SOX6 ChIP-seq confirmed direct SOX6 binding at regulatory elements associated with genes controlling OPC morphogenesis, migration, and glutamatergic signaling. CUMS reduced SOX6 protein levels and SOX6-associated regulatory network activity in OPCs, whereas OPC-specific enhancer-driven restoration of SOX6 rescued stress-induced defects in OPC migration and myelination. Together, these findings define a stress-sensitive SOXD/SOX6 regulatory mechanism linking chronic stress to oligodendrocyte lineage dysfunction and myelination deficits, identifying SOX6 as a functional regulatory node with therapeutic potential for stress-related brain pathology.

neuroscience↗

bioRxiv2026

Microbial Dynamics Across Commercial Spaceflights of Varying Duration

SummarySpaceflight introduces environmental stressors that can alter human microbiomes and immune responses. We analyzed 259 biospecimens from six astronauts across two commercial ISS missions: Axiom 2 (10-day mission) and Axiom 3 (21-day mission). Samples included saliva, stool, urine, and body swabs from 10 anatomical sites, profiled via 16S ribosomal RNA (rRNA) gene sequencing. Gut and oral microbiomes remained stable, while skin-associated communities exhibited transient diversity shifts post-flight. Taxonomic analysis revealed individual and site-specific patterns as well as a possible microbial acquisition from the ISS/space-flight environment and microbiome exclusivity. Cytokine profiling from single cell data indicated immune activation, with IL-32 and IL-16 elevated in Axiom 2 and Axiom 3, coinciding with microbial changes. These findings provide an integrated view of microbiome individuality, exclusivity and immune dynamics during two short-duration commercial spaceflights of three weeks, informing strategies for crew health on future long-duration missions.

microbiology↗

bioRxiv2026

Toward CT-based Tractography: Presurgical White Matter Tract Mapping in Intracerebral Hemorrhage

Presurgical mapping of key white matter (WM) fiber tracts is crucial for intracerebral hemorrhage (ICH) surgery, but it currently relies on tractography from diffusion MRI (dMRI), which has limited applicability in urgent or resource-constrained settings due to long scan times and limited MRI availability. To bridge this gap, we developed a deep learning approach designed to reconstruct critical fiber tracts directly from routinely acquired CT scans, focusing on the corticospinal tract (CST) due to its high clinical relevance. By training a novel network on a curated dataset of 150 paired CT-dMRI scans (101 ICH patients and 49 healthy subjects), we enabled the direct mapping of the CST from CT images alone, bypassing the traditional requirement for dMRI. Our results demonstrate that the CT-derived tracts achieve high anatomical plausibility, with neurosurgeon expert assessments yielding a Likert score of 3.64. Furthermore, the clinical relevance of these reconstructions was validated by a significant correlation between CT-derived tract integrity and patient motor scores (r = 0.726, p = 3.731x 10-7). These findings suggest that in complex clinical scenarios, particularly where dMRI signal quality is compromised by lesion-induced distortion, this CT-based mapping may serve as a useful anatomical reference. Overall, by enabling CT-based WM mapping, our method potentially offers a practical route to gain the key advantages of tractography in resource-limited or time-critical settings

neuroscience↗

bioRxiv2026

Accurate protein stability prediction for small domains using mega-scale experiments

Predicting absolute protein folding stability is a long-standing challenge in biophysics, with broad applications in protein design and in understanding genetic variation and evolution. Physics-based simulations have shown limited success at predicting stability and are often computationally intractable, and machine learning methods have been constrained by the lack of sufficiently large experimental datasets. We recently introduced cDNA display proteolysis, a cell-free approach that can measure folding stability for nearly one million protein domains in parallel. Here, we applied this method to measure stability for 1.8 million diverse protein domains 60-80 amino acids in length primarily taken from the MGnify metagenomic database and spanning over 200,000 sequence families. Using this new "MGnify Stability dataset", we developed the predictive models SaProt{Delta}G and ESM3{Delta}G, which accurately predict absolute folding stability for small domains with root mean squared error of 0.8 kcal/mol over a 6 kcal/mol range (Spearman rank correlation of 0.88). These predictors show high accuracy at predicting effects of substitutions, insertions, and deletions, successfully identify global trends toward higher stability in thermophilic organisms, and improve discrimination of stable and unstable computationally designed proteins. Our results illustrate how megascale biophysical measurements can complement existing evolutionary and structural data to enable accurate absolute stability prediction for small domains.

biophysics↗

bioRxiv2026

The microstructure-weighted human connectome: network properties and structure-function correlations across spatial scales

Conventional connectome edge weights, such as number of streamlines (NOS) or diffusion tensor imaging (DTI) metrics, lack specificity to microstructural details which may hold relevance for macroscale brain organisation. Since biophysical diffusion modelling offers greater specificity to microstructure, we investigated whether parameters from the Standard Model of white matter diffusion provide informative alternatives for connectome weights - namely the intra-axonal signal fraction (f) and perpendicular extra-axonal diffusivity (DePerp). Using diffusion MRI data from healthy adults, we constructed structural networks at four parcellation scales, weighted by f, DePerp, NOS, fractional anisotropy (FA) and radial diffusivity (RD). While all weights reproduced expected small-world properties and hub structure, only DePerp and normalised NOS captured non-random properties of local organisation at all spatial scales. We then correlated each weighted connectome with resting-state fMRI functional connectivity and intracranial measurements of conduction velocity. At both the whole-brain and regional level, NOS gave strong coupling with fMRI functional connectivity. Connectomes weighted by DePerp, FA and RD gave weak but significant coupling with fMRI functional connectivity and conduction velocity. Notably, DePerp and RD gave highest consistency in regional structure-function coupling across spatial scales. Thus, connectome weights derived from DePerp capture meaningful aspects of brain network organisation with functional relevance.

neuroscience↗

bioRxiv2026

Nature-inspired nanodiscs for lesion-targeted delivery reprogram macrophages and attenuate established abdominal aortic aneurysms

Abdominal aortic aneurysm (AAA) is a life-threatening vascular disease characterized by chronic inflammation and immune dysregulation, with lesional macrophages playing a pivotal role in disease progression. However, effective and safe delivery of immune modulators to macrophages at the site of AAA remains a major clinical challenge. To address this unmet need, we report a nature-inspired nanodisc platform based on high-density lipoproteins for targeted delivery to lesional macrophages, further engineered with a multi-component targeting strategy incorporating an aneurysm-homing peptide and phosphatidylserine lipids. Nanodiscs encapsulating an anti-inflammatory protein kinase R-like endoplasmic reticulum kinase (PERK) inhibitor remarkably attenuated progression of established AAA in an elastase-induced mouse model. Using a combination of in vivo biodistribution and immune profiling approaches, we demonstrate that nanodisc-assisted PERK inhibitor delivery selectively reprograms the local immune microenvironment and attenuates pathological inflammation in AAA disease models. Notably, a single administration achieves sustained therapeutic efficacy with favorable safety profiles, effectively limiting the progression of established AAA in a clinically relevant setting. This work presents a new avenue of designer nanomedicines for targeted immunomodulation and maybe broadly applicable for a wide range of vascular and immune-mediated pathologies.

bioengineering↗

bioRxiv2026

Human-like sequential sound-to-meaning transfer drives artificial speech comprehension

Artificial intelligence has reached a pivotal threshold. Multimodal large models can approach human-level speech comprehension by rapidly transforming sound into meaning. However, whether this process relies on human-like mechanisms remains unknown. Here, we compared the human brain with twelve speech language models (SLMs) using a phonology-semantics confusion paradigm. Stereo-electroencephalography revealed two mechanisms of phonology-to-semantics (P2S) transfer in the human brain: a local sequential transformation within specific neuronal populations, and a global cross-regional hierarchy of P2S representations. Only brain-model alignment in the local sequential manner predicted model performance. Correspondingly, targeted lesioning of local sequential P2S-transfer model units markedly impaired comprehension performance, while activation steering of these units improved performance. In addition, such local sequential P2S-transfer model units were identified across languages. Together, this study establishes local sequential P2S transformation as a fundamental computational principle shared across biological and artificial intelligence, offering a mechanistic bridge for future brain-inspired speech systems.

neuroscience↗

bioRxiv2026

Near-Infrared-II Chemo-Optogenetics for Deep-Brain Stimulation

Precise activation of ion channels enables fine-tuned control of neuronal excitability, providing a powerful strategy for dissecting and modulating neural circuits. Although optogenetics offers high spatiotemporal neuronal manipulation with cell-type specificity, its reliance on visible light (400-650 nm) limits tissue penetration to millimeter depths, restricting applications in deep-brain stimulation. Here we report a near-infrared-II (NIR-II)-sensitive calcium channel, HaloNeu, created by genetically fusing a circularly permuted HaloTag (cpHaloTag) to the thermo-sensitive transient receptor potential vanilloid 1 (TRPV1) and covalently conjugating NIR-II photothermal nanotransducers (HPN). The HaloNeu enables non-invasive neuromodulation at depths up to 1.0 cm at ultralow laser power ([~]60 mW/cm2) and up to 5.0 cm under the safe exposure limit ([~]1 W/cm2) with 1064 nm laser illumination. Remarkably, HaloNeu maintains stable, on-demand neuron-specific modulation for over two months in vivo, providing sustained activation of ventral tegmental area (VTA) circuits and effective alleviation of Parkinsonian symptoms in mouse models. These results establish HaloNeu as a robust and versatile platform for cell-type-specific, deep-tissue, and chronic neuromodulation, with broad implications for neuroscience and neurotherapeutics.

neuroscience↗

bioRxiv2026

TopoFuseNet: Hierarchical Graph Representation Learning with Multi-Scale Topological Features for Accurate Drug Synergy Prediction

Accurate prediction of drug synergy is paramount for developing effective combination therapies and advancing personalized medicine. Although methods based on graph neural networks (GNNs) have become a prevalent approach, they often treat molecules as flat graphs of connected atoms, thus overlooking their inherent hierarchical structure (i.e., atoms forming functional groups) and the critical topological information that governs molecular interactions. To address this limitation, we introduce TopoFuseNet, a novel hierarchical graph representation learning framework that integrates multi-scale topological features. The core innovations of TopoFuseNet include: 1) The first-ever application of "Group Centrality" from network science to cheminformatics, enabling the identification and quantification of functional groups crucial to drug activity; 2) A systematic, multi- path strategy to seamlessly integrate node-level (atom) and group-level (functional group) topological features into a Graph Attention Network (GAT) via feature augmentation, attention biasing, and hierarchical pooling; 3) A Differential Transformer module to deeply fuse multi-modal features learned from sequences, fingerprints, and our proposed hierarchical graph representations. Extensive experiments on two large-scale benchmark datasets, DrugComb and DrugCombDB, demonstrate that TopoFuseNet significantly outperforms state-of-the-art methods across multiple key metrics, including AUC, AUPRC, and F1-score, while exhibiting exceptional generalization robustness under various stringent cold-start scenarios. In-depth ablation studies further confirm the effectiveness and necessity of each proposed innovative module. Furthermore, multi-scale interpretability analysis and zero-shot cross-domain transfer experiments reveal that the model successfully captures molecular interaction rules with clear pharmacological significance, demonstrating immense practical potential for discovering novel combination therapies through large-scale virtual screening. Our work not only delivers a superior model for drug synergy prediction, but more importantly, it establishes a novel and scalable paradigm for effectively integrating hierarchical molecular structures and topological information into GNNs.

bioinformatics↗

bioRxiv2026

Functional Landscape of Motifs within the Sarbecovirus Spike Cytoplasmic Tail

Sarbecoviruses exhibit extensive diversity in host range and zoonotic potential. Although the ectodomain of the viral spike protein has been well characterized, the functional landscape of the cytoplasmic tail (CT) remains poorly defined. To address this gap, we systematically generated CT truncation variants of representative spike proteins from all major ACE2-utilizing clades to define the roles of conserved host factor-interacting motifs associated with COPI, COPII, FERM, and SNX27. Using vesicular stomatitis virus (VSV)- and lentivirus-based pseudotyping systems, we evaluated viral entry in cells expressing varying levels of ACE2 and TMPRSS2. Our results demonstrate that CT-associated motifs differentially regulate viral infectivity. Specifically, truncation of the COPI- or SNX27-binding motifs markedly reduces entry efficiency, whereas disruption of the COPII-binding motif produces the opposite outcome. By contrast, removal of the FERM-binding motif consistently enhances infectivity across lineages. Mechanistically, truncation of this motif increases spike expression, cell surface localization, incorporation into virions, and particle stability. Importantly, despite these pronounced effects on viral infectivity, deletion of the FERM-binding motif does not affect antigenicity, receptor dependence, or sensitivity to protease inhibitors, as demonstrated by neutralization and inhibition assays. In addition, this approach substantially increases spike protein density on virus-like particles (VLPs). Collectively, by extending the analysis beyond SARS-CoV-1 and SARS-CoV-2, our study reveals a generalizable mechanism in which cytoskeletal anchoring mediated by the FERM-binding motif acts as a limiting determinant of viral assembly. These findings provide a practical framework for optimizing pseudovirus platforms and guiding vaccine development against emerging viral threats.

microbiology↗

bioRxiv2026

IFITM1 inhibits Henipavirus membrane fusion by trapping ephrinB2 receptors in fusion-unfavorable membrane nanodomains.

Nipah (NiV) and Hendra (HeV) viruses are prototype henipaviruses (HNV) within the family of Paramyxoviridae. They are highly pathogenic zoonotic paramyxoviruses that cause fatal encephalitis in humans, with case fatality rates of 40-70%. These viruses use the receptor-binding protein (G) to engage host receptors ephrinB2 and/or ephrinB3, while the fusion protein (F) drives membrane fusion leading to virus entry and cell-cell fusion. Interferon-inducible transmembrane (IFITM) proteins restrict the infection of many unrelated viruses at virus entry. We show that the knockdown of IFITM1 enhances HNV entry and cell-cell fusion in both human epithelial and endothelial cells, while overexpression of IFITM1 inhibits both. Using single-molecule localization microscopy, we observed that IFITM1 forms nanoclusters on the plasma membrane that are enhanced upon interferon stimulation. IFITM1 clusters spatially co-occur with GM1-enriched, raft-like membrane domains. Notably, IFITM1 interacts with ephrinB2, and redistributes ephrinB2 into larger and denser clusters, potentially within GM1-enriched regions. Using single-particle tracking, we further show that ephrinB2s lateral diffusion on the plasma membrane is confined by IFITM1, and this confinement is not reversed by amphotericin B, a membrane-permeabilizing agent that increases membrane fluidity. Our data suggest that IFITM1 sequesters ephrinB2 in fusion-unfavorable membrane domains, thereby enhancing the energy barrier for virus-cell membrane fusion and decreasing the virus-receptor encounters. SignificanceNipah (NiV) and Hendra (HeV) viruses are zoonotic paramyxoviruses that cause fatal encephalitis in humans. Here, we show that IFITM1 inhibits NiV and HeV entry at the surface of epithelial and endothelial cells. We report that IFITM1 restricts NiV entry by sequestering the host receptor ephrinB2 in membrane domains that are energetically unfavorable for virus-cell membrane fusion. IFITM1 also restricts ephrinB2 lateral diffusion on the plasma membrane, thereby reducing productive virus-receptor encounters. Notably, amphotericin B treatment, which increases membrane fluidity, did not rescue IFITM1-mediated confinement of ephrinB2 or restore NiV entry. Our data reveal a new mechanism by which IFITM1 restricts NiV infection, suggesting that its antiviral activity extends beyond simply altering membrane fluidity.

microbiology↗

bioRxiv2026

EnzCast: Prediction of Patient-Specific Enzymatic Kinetics through Multi-Modal Deep Learning and Isoform-Resolved Bayesian Inference based on Single-Cell Transcriptomics

Enzyme kinetic parameters underpin mechanistic biology but remain sparse in physiological context. We present EnzCast, a multi-modal framework jointly predicting Km, kcat, kcat/Km, and Ki from protein sequence, 3D structure, substrate chemistry, and experimental conditions, paired with IsoKin, an isoform-resolved Bayesian framework converting EnzCast priors into patient-specific in vivo kinetics. Trained on KinBench, the largest curated kinetics database, task-adaptive EnzCast achieved R2 = 0.413, 0.455, 0.227, and 0.105 for Km, Ki, kcat, and kcat/Km, surpassing all baselines on catalytic tasks. Systematic condition scans recovered compartment-specific pH direction inversion and pathway-level temperature responses. In a 20-patient colorectal cancer single-cell cohort, IsoKin reduced posterior uncertainty by 73.3% and 77.3%, revealing cell-type-specific rewiring. Orthogonal validation--scFEA flux, DepMap essentiality (permutation P = 0.0008) and TCGA survival-- provided mixed but directionally consistent support. Together, EnzCast and IsoKin bridge in vitro prediction, condition-aware biochemical interrogation and patient-resolved in vivo inference.

bioinformatics↗

bioRxiv2026

MVCBench: A Multimodal Benchmark for Drug-induced Virtual Cell Phenotypes

Drugs induce coordinated phenotypic changes across multiple modalities, including transcriptional reprogramming and cellular morphological remodeling. Predicting these drug-induced modality changes is central to drug discovery, mechanism-of-action studies and precision therapeutics, however, prediction performance depends critically on how both drug compounds and cellular states are represented. Despite rapid advances in drug molecular and gene representation methods, a systematic evaluation of these methods remains lacking. Herein, we introduce MVCBench, a comprehensive benchmarking framework for evaluating drug molecular and gene representation methods in predicting drug-induced multimodal virtual cell (MVC) phenotypes. MVCBench leverages large-scale transcriptomic and high-content imaging data and systematically evaluates 24 representation methods (12 drug molecular and 12 gene representation methods) across nearly 1.1 million drug-induced profiles, under both in-distribution and out-of-distribution settings spanning unseen compounds, cell lines, assay plates and datasets. Our benchmarking reveals a pronounced modality-dependent asymmetry: advanced drug molecular representations substantially improve the prediction of drug-induced morphological phenotypes but provide only limited gains for gene expression prediction relative to classical fingerprints, whereas task-specific gene representations outperform general-purpose foundation models in predicting drug-induced transcriptomic responses. Predictive performance also deteriorates sharply under distribution shift, highlighting persistent challenges in cross-dataset and cross-platform generalization. We further show that integrating transcriptomic and morphological modalities consistently improves prediction accuracy, and derive practical design principles for MVC architectures, including modality-aware loss calibration and fusion strategies. Together, MVCBench provides a systematic foundation for evaluating representation methods and offers guidance for developing robust MVC models of drug-induced cellular responses.

bioinformatics↗