Biology subjects

Li, Q.

Publications and source records attributed to Li, Q..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

PalmLab: A Comprehensive Computational Platform for Systematic Annotation and Functional Interrogation of Protein Palmitoylation

Protein palmitoylation is a dynamic and reversible post-translational lipid modification with progressively recognized pathophysiological relevance in cancer and a range of human diseases. Current databases, however, predominantly catalog palmitoylation sites derived from isolated literature reports, and suffer from a conspicuous lack of systematic harmonization across publicly available mass spectrometry (MS)-based palmitoylome datasets, which severely impedes standardized annotation and cross-cohort comparative analyses. Here, we introduce PalmLab, an integrative computational platform that synergistically couples MS-based palmitoylome data curation with interactive, multi-dimensional analytical functionalities for both human and mouse proteomes. Through a unified and rigorously standardized bioinformatics pipeline, PalmLab enables systematic reanalysis of all published palmitoylome datasets, yielding a curated compendium of 15,968 human and 9,924 mouse MS-validated palmitoylated proteins, a quantitative increase that more than doubles the total entries available in existing public repositories. The platform further provides user-friendly, code-free analytical modules that support differential palmitoylation profiling, context-specific pattern visualization, protein-protein correlation network inference, mutation-palmitoylation association mapping, and sequence-based motif discovery. Collectively, PalmLab constitutes a comprehensive and openly accessible resource for systematic palmitoylome mining and functional interrogation of palmitoylation events, thereby facilitating the elucidation of palmitoylation-mediated regulatory circuits and aiding the prioritization of novel therapeutic targets in oncology. The platform is freely available at https://palmlab.intelligent-oncology.com.

bioinformatics↗

bioRxiv2026

scRep: A Latent-Space Self-Distilled Foundation Model for Single-Cell Representation Learning

Single-cell foundation models have shown strong potential for learning transferable representations from large-scale transcriptomic data. However, many existing approaches rely on reconstructing masked gene expression values, creating a potential mismatch between observation-space reconstruction and the goal of learning stable biological representations. This challenge is particularly relevant to single-cell RNA sequencing, where sparsity, incomplete gene detection, and technical variation can obscure the underlying biological state. Here, we introduce scRep, a compact latent-space self-distillation framework for single-cell representation learning. Rather than reconstructing raw expression values, scRep aligns differently perturbed views of the same cell through a momentum-updated teacher--student architecture, with self-distillation objectives at both the cell and gene levels. This representation-centered formulation encourages the model to capture biological information that remains stable across incomplete and perturbed transcriptomic observations. Using frozen representations without task-specific fine-tuning, scRep pretrained on approximately 2.8 million cells achieves the strongest overall performance across the evaluated frozen-representation benchmarks, demonstrating strong sample efficiency. A larger-scale scRep model pretrained on 30.72 million cells further demonstrates that the framework remains effective when scaled to a substantially larger and more diverse corpus. Beyond cell identity, scRep prioritizes established marker genes, recovers transcription factor--associated gene programs with cell-type-specific activity, and preserves continuous developmental structure that supports graph-based pseudotime inference. We further show that pretraining performance is closely associated with biological diversity: reducing redundant cells while improving cell-type coverage can match or exceed the performance of larger, less balanced training corpora. Together, these results establish latent-space self-distillation as an effective alternative to expression reconstruction for single-cell foundation modeling and suggest that efficient scaling depends not only on the number of cells, but also on the learning objective and the biological diversity of the pretraining corpus.

bioinformatics↗

bioRxiv2026

A Hymenoptera-restricted gene mediating ant castes co-opts deeply conserved machinery to control organ size

Lineage-specific genes are widespread and have been implicated as phenotypic innovation inducers, but how they acquire complex developmental functions remains poorly understood. Ant queens and workers develop dramatically different organ sizes from identical genomes under juvenile hormone (JH) control, yet the molecular effectors translating JH signalling into caste-specific organ growth remain unknown. Here we identify torch, a Hymenoptera-restricted gene, as the most consistently gyne-biased and JH-responsive gene across 68 ant species. Knockdown of torch in virgin queens of Monomorium pharaonis produces a worker-like, multi-organ growth-restricted phenotype. Mechanistically, torch harbours an E-box-like motif activated by the JH receptor Gce-Tai and acts as a GA-repeat-binding transcription factor that regulates Hippo signalling, the deeply conserved organ-size control pathway in animals. Expressing torch heterologously in mice and a growth-restricted Drosophila background shows that the gene retained its general growth-promoting activity across more than 700 million years of animal evolution in lineages that lack the gene, establishing that its function is mediated through conserved rather than ant-specific machinery. A lineage-specific gene can therefore acquire complex morphogenetic function by co-opting ancient organ-size circuitry, providing a general route by which novel genes can drive phenotypic innovation.

evolutionary biology↗

bioRxiv2026

Beyond Imbalance: An Elasticity Framework for the Distance-averaged Force-Velocity Relationship in Vertical Jump

This study aimed to (1) establish the distance-averaged F-V relationship framework and (2) develop elasticity metrics that quantify how F-V relationship variables govern jump height and inform training prescription. Theoretical derivation and experimental validation across 108 F-V relationship models derived from 1578 jumps (countermovement jump and squat jump at three knee angles; 20 well-trained subjects) yielded a standard error of 2.1% and a nearly perfect correlation (r = 0.96, p < 0.001) between measured and predicted jump height. Four elasticity metrics were formulated: force elasticity (F_{e}), the elasticity of jump height to maximal force (F_{0}); velocity elasticity (v_{e}), the elasticity of jump height to maximal velocity (v_{0}); the force-velocity elasticity norm {(\mathrm{F}-\mathrm{V}}_{\mathrm{EN}}=\sqrt{F_{e}^{2}+v_{e}^{2}}), reflecting the overall sensitivity of jump height to changes in F-V relationship variables; and the force-velocity elasticity ratio {(\mathrm{F}-\mathrm{V}}_{\mathrm{ER}}=F_{e}{\div v}_{e}), indicating which variable dominates the jump height response. Simulations and experiments revealed that F_{e} bore an inverse relationship to F_{0}, and v_{e} was inversely related to v_{0}, reflecting diminishing marginal returns. At a fixed jump height, simulations showed {\mathrm{F}-\mathrm{V}}_{\mathrm{EN}} and {\mathrm{F}-\mathrm{V}}_{\mathrm{ER}} displayed a U-shaped relationship; a balanced profile ({\mathrm{F}-\mathrm{V}}_{\mathrm{ER}}=1) did not always correspond to the lowest {\mathrm{F}-\mathrm{V}}_{\mathrm{EN}}. The distance-averaged F-V elasticity framework offers a physically grounded and quantitative tool for linking F-V relationship variables directly to jump performance, providing a basis for informing individualized training decisions.

biophysics↗

bioRxiv2026

Quantifying sprint force-velocity elasticity: implications for individualized training decisions

This study aimed to (1) develop an elasticity framework for the sprint force-velocity (F-V) relationship and (2) examine how maximal force (F_{0}), maximal velocity (v_{0}), and sprint distance modulate the four derived elasticity metrics, and (3) explore these elasticity metrics' interrelation. After modelling the F-V relationship differential equation, four elasticity metrics were defined as force elasticity (F_{e}), the elasticity of sprint time to F_{0}; velocity elasticity (v_{e}), the elasticity of sprint time to v_{0}; the force-velocity elasticity norm {(\mathrm{F}-\mathrm{V}}_{\mathrm{EN}}=\sqrt{F_{e}^{2}+v_{e}^{2}}), capturing the combined sprint time sensitivity to proportional changes in F_{0} and v_{0}; and the force-velocity elasticity ratio {(\mathrm{F}-\mathrm{V}}_{\mathrm{ER}}=F_{e}{\div v}_{e}), indicating which variable dominates the sprint time response. Model simulations showed that F_{e} decreased with rising F_{0} and increased with rising v_{0}, while v_{e} showed the opposite pattern. With increasing sprint distance, F_{e} decreased and v_{e} increased. Given its negligible effect on sprint time, ignoring air resistance yields a conservation law (2F_{e}+v_{e}\equiv 1), indicating that a gain in one elasticity metric necessarily diminishes the other in a fixed proportion. This framework also identifies a valley distance (d_{valley}) at {\mathrm{F}-\mathrm{V}}_{\mathrm{ER}}=2, where {\mathrm{F}-\mathrm{V}}_{\mathrm{EN}} is minimized (\sqrt{0.2}) and sprint time is least responsive to changes in F-V relationship variables. Empirical data confirmed that the two theoretical laws still hold approximately when air resistance is considered. By linking changes in F_{0} and v_{0} to sprint time across different distances, the elasticity framework provides a quantitative basis for estimating the theoretical sprint time response to documented changes in F-V relationship variables.

biophysics↗

bioRxiv2026

A closed-loop reinforcement learning framework for rapid compound directed optimization

Generative artificial intelligence (AI) holds transformative potential for drug discovery, yet existing architectures typically operate in open loops without experimental feedback. Here we introduce rapid compound directed optimization (RCDO), a closed-loop reinforcement learning framework that accelerates the optimization process by bridging dry-lab computation with wet-lab feedback. RCDO couples a three-dimensional structure-guided generative model with a multi-level reward system updated after each design cycle using experimental measurements from all synthesized compounds, including inactive or developability-failed compounds. By continuously aligning the generative model with accumulated wet-lab measurements, RCDO substantially compresses optimization timelines. We evaluated RCDO through retrospective benchmarking against historical optimization trajectories and prospective wet-lab campaigns targeting ROR1, NLRP3, and NSD3. Across prospective evaluations, RCDO rapidly resolved key optimization bottlenecks within two to three design cycles: improving the oral exposure of an ROR1 inhibitor by 40-fold while maintaining antitumor efficacy, reducing CYP2C19 inhibition of an NLRP3 antagonist by 20-fold while preserving inflammasome activity, and boosting the binding affinity of an NSD3 hit by 18-fold. By directly coupling wet-lab feedback to generative learning, RCDO establishes an efficient platform for compound directed optimization, transforming AI-driven drug discovery from static generation into continuous experimental adaptation.

bioinformatics↗

bioRxiv2026

DeMoP: A Language-Model-Guided Mixture-of-Experts Framework for Cancer Prognosis

Integrating heterogeneous clinical and molecular data for cancer prognosis remains challenging because their dimensionality, semantics and distributions differ across patients and cohorts. Here we present DeMoP, a language-model-guided mixture-of-experts framework that serializes structured patient profiles as natural-language sequences and learns adaptive prognostic representations from clinical variables, copy-number alterations, and gene descriptions. DeMoP combines a fine-tuned DeBERTa-v3-large encoder, attention-based token pooling, and a residual mixture-of-experts prediction head. In held-out tests from two independent pan-cancer cohorts, GENIE (63,090 patients) and TCGA (4,123 patients), DeMoP outperformed the conventional machine-learning and deep-learning baselines evaluated, achieving AUROCs of 0.939 and 0.805 and class-1 F1 scores of 0.72 in both cohorts. A GENIE-trained model transferred directly to TCGA with an overall class-1 F1 score of 0.62. Gene-level ablations recovered established cancer-associated genes and highlighted less-studied candidates. DeMoP provides a unified approach to heterogeneous biomedical data integration, cross-cohort outcome prediction, and model interpretation.

bioinformatics↗

bioRxiv2026

Connectome-constrained modeling identifies neurons and synapses that sustain spontaneous activity in Drosophila

Synapse-resolution connectomes specify a brain's wiring; brain-wide recordings capture its activity. Neither alone identifies the cells and synapses that generate the activity. We bridge them by fitting a FlyWire connectome-constrained whole-brain model to calcium recordings of spontaneous activity in head-fixed Drosophila, then probing it in silico at cellular and synaptic resolution. The fitted model reproduces three features it was never trained on: lognormal synaptic weights, scale-free neuronal avalanches, and short intrinsic time constants in visual cells, each consistent with experiment. Systematic perturbations of the digital whole-brain model show that spontaneous activity is not distributed uniformly across the connectome, but is organized by a compact neuropil core. Within this core, a highly sparse, brain-spanning ensemble of inhibitory hub neurons and their reciprocal synapses with excitatory partners are necessary and sufficient to sustain whole-brain resting-state dynamics. Connectome-constrained modeling therefore converts wiring diagrams and recordings into a perturbable digital platform that identifies the cells and synapses sustaining resting-state dynamics.

neuroscience↗

bioRxiv2026

Specification of embryonic shoot stem cells via a small RNA-driven morphogenic circuit

The specification of embryonic stem cells capable of self-renewing and differentiation into virtually any cell type, is one of the most consequential events in the development of a multicellular organism. Yet the mechanisms establishing embryonic stem cell fate remain poorly understood, particularly in monocotyledonous cereals. Using the classic mutant leafbladeless1-raggedseedling1, we show that the small RNA tasiARF acts as a primary epidermis-derived morphogenic signal that organizes shoot stem cell specification in the maize embryo. tasiARF restricts expression of the AUXIN RESPONSE FACTOR 3 (ARF3) transcription factor, which modulates cell wall mechanics and guides the differential localization of PIN-FORMED (PIN) auxin efflux carriers, establishing a localized auxin minimum permissive for stem cell fate. Interestingly, loss of this auxin minimum and the associated shoot stem cell defects in tasiARF-deficient embryos are buffered by natural variation at a quantitative trait locus (QTL) controlling expression of MICROTUBULE-ASSOCIATED PROTEIN 65-3 (MAP65-3), a critical factor determining cell division orientation, which reshapes auxin dynamics and restores stem cell specification, and is itself under tasiARF-ARF3 control. Thus, embryonic shoot stem cell specification in maize is governed by an intricate morphogenic circuit that couples small RNA-mediated positional information to a self-stabilizing network interdependently linking cell wall mechanics, auxin signaling, and cell division patterning. This mechanistic framework reveals the redeployment of an ancient small RNA pathway as a lineage-specific innovation to establish a conserved, stem cell-permissive low auxin environment within the divergent embryonic architecture of monocotyledonous cereals. More broadly, it identifies molecular entry points for the engineering of embryogenic competence and regeneration capacity for the improvement of cereal crops.

plant biology↗

bioRxiv2026

Proteome-Scale Mining and Multi-Objective Prioritization of Encrypted Antimicrobial Peptides with Experimental Validation

Encrypted antimicrobial peptides (eAMPs) are bioactive fragments embedded within larger proteins and represent an underexplored source of antimicrobial candidates. We developed a multi-layer proteome-mining framework to identify and prioritise eAMPs from 95%-identity-reduced protein sets derived from 265 high-quality bacterial genomes. Three complementary, layer-specific extraction strategies targeting protein termini, internal cleavage sites, and cationic hotspots yielded 29,251,180 unique peptide candidates. Dual AMP prediction with AMP-scanner v2 and Macrel reduced this space to 3,249,772 consensus candidates. Downstream prioritisation followed two complementary routes: a low-haemolysis branch focused on selectivity-oriented candidates and a high-activity branch that retained predicted haemolytic sequences as mechanistic comparators. Structure prediction and review were performed for 185 candidates, and 18 entered Tier-1 developability, novelty, and membrane-activity assessment. Three sequence-matched representatives were selected for experimental evaluation. Molecular-dynamics simulations supported water-phase stability of GEAMP_71c139393ac596b5 and deep anionic-membrane insertion by GEAMP_12ffb5d589c8cb1b. In replicated colony-count assays against Escherichia coli and Staphylococcus aureus, all three peptides showed concentration-dependent activity over 8-128 uM. GEAMP_12ffb5d589c8cb1b was the most active, producing 1.52- and 2.27-log10 reductions, respectively, at 128 uM relative to the matched 8 uM condition. Together, these results establish a sequence-traceable workflow linking proteome-scale eAMP discovery with structural prioritisation and experimental activity assessment.

bioinformatics↗

bioRxiv2026

LNP-CpG: deploy the self-adjuvant role of mRNA vaccines

With the rapid advancement of mRNA vaccines, lipid nanoparticles (LNPs) have emerged as pivotal carriers and adjuvants for non-mRNA vaccine modalities, driven by their superior nucleic acid delivery efficiency and intrinsic self-adjuvanting properties. In this study, we systematically evaluated various formulation strategies combining LNPs and the CpG adjuvant within a varicella-zoster virus glycoprotein E (VZV-gE) subunit vaccine framework. We demonstrated that uniform nanoparticles formed by LNP-encapsulated CpG (LNP-CpG), when simply admixed with the gE antigen, elicited superior immunogenicity compared to alternative encapsulation configurations. Intramuscular administration of a two-dose (LNP-CpG)+gE regimen significantly augmented both humoral and cellular immune responses in mice, markedly outperforming the commercial vaccine Shingrix (administered at a 1/10 human dose). Crucially, the identical regimen induced robust, comparable immune profiles to a full human dose of Shingrix in rhesus macaques. Furthermore, LNP-CpG displayed broad-spectrum utility across diverse vaccine platforms, demonstrating efficacy against both respiratory (RSV) and neurotropic (HSV) pathogens, compatibility with multiple modalities, including subunit (VZV-gE, RSV-Pre-F), live-attenuated (LA-HSV), and inactivated (i-HSV) vaccines; and versatile implementation in a combined VZV+RSV formulation. Collectively, our findings position LNP-CpG as a versatile, safe, highly efficacious adjuvant platform with substantial clinical translational potential, offering a compelling paradigm for next-generation vaccine development.

immunology↗

bioRxiv2026

Genome-context-aware discovery of antibacterial peptides from bacterial small open reading frames

Small open reading frames (sORFs) are a potentially rich, yet error-prone, source of antimicrobial-peptide (AMP) candidates: short sequences are readily prioritized by AMP classifiers but may derive from incomplete gene calls. We developed a genome-context-aware discovery workflow that separates AMP-like sequence properties from evidence for a complete, recurrent coding locus. From 649,653 RefSeq assemblies representing 327 clinically relevant bacterial species, species-aware clustering and length filtering yielded 4,442,548 representative 10-100-aa sequences. AmpScanner v2, Macrel and AMPlify identified 585 non-haemolytic records supported by all three models. However, genome-context auditing of 11,918 mapped candidates showed that 529 of 536 mapped consensus candidates were supported exclusively by partial ORFs near contig termini. By contrast, 3,382 candidates had at least one complete non-edge occurrence; 1,069 recurred in [≥]2 assemblies and 251 in [≥]10 assemblies. We therefore assembled a 20-peptide panel through two explicitly labelled routes: sequence/structure-led selection (n=8) and genome-supported selection (n=12). Broth microdilution against Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 25923 identified low-micromolar activity in both routes. CAND_04141, a recurrent complete non-edge candidate, had the strongest combined profile (MICs of 4 and 2 M, respectively), while CAND_07825 and CAND_04265 were also active at low micromolar concentrations. In plate-count MBC assays, all three advanced peptides achieved [≥]3-log10 reductions at 128 M. These findings show that high classifier agreement is not a substitute for genomic evidence and provide an auditable framework for prioritizing both synthetic AMP-like sequences and candidate genome-encoded peptides.

bioinformatics↗

bioRxiv2026

Population genomics of the inquiline social parasite Acromyrmex insinuator and its leaf-cutting ant hosts A. echinatior and A. octospinosus reveals cryptic differentiation and reduced efficiency of selection in the parasite

Inquiline social parasites usurp colonies of closely related host ants to exploit their social resources. They are almost invariably rare, patchily distributed and difficult to study. Here we build on almost 25 years of Panamanian fieldwork on the social parasite Acromyrmex insinuator and its A. echinatior and A. octospinosus hosts, to perform a population genomic analysis to test hypotheses that have been suggested to shape the evolution of inquiline social parasites: 1. Do these parasites indeed have extremely reduced effective population size? 2. Does extant genetic variation at coding and non-coding sites carry signatures of erosion of adaptive potential? 3. Has A. insinuator become fully reproductively isolated from its sympatric hosts and how closely related are its primary and secondary host? We show that the two host species are completely distinct and that genetic diversity and effective population size of the social parasite are dramatically reduced despite ongoing but very minor recent gene flow between the parasite and its primary host A. echinatior. We also demonstrate that non-synonymous codon-sites evolved at rates nearly indistinguishable from synonymous codon-sites. This indicates a significant reduction in the efficiency of natural selection consistent with inquiline social parasite lineages generally being evolutionarily short-lived. We finally uncover clear sub-structure in the parasite population, with two genetically distinct lineages occurring in sympatry in the Panama Canal Zone, and with significant differences in their likelihood of exploiting the secondary host A. octospinosus and the primary host from which they segregated sympatrically ca. 1 MYA.

genomics↗

bioRxiv2026

Conserved influenza A epitope candidate regions and a benchmark of ESM-2 sequence features

Influenza A virus antigenic drift forces annual vaccine reformulation, motivating the search for conserved epitope candidates that could support broadly protective vaccines. We systematically screened influenza A virus sequences (H1N1, H3N2, H5N1; nine viral proteins) to define 98 conserved candidate regions, 38 of which were identical across the H1N1, H3N2, and H5N1 consensus sequences, all in the polymerase complex and nucleoprotein (PB2, PB1, PA, NP), whereas the ten surface-glycoprotein (HA/NA) candidates were subtype-specific. We then benchmarked two protein-language-model (ESM-2) features against alignment conservation. Group-masked log-probability correlated moderately with MSA conservation (Spearman rho = 0.25 to 0.39 for HA) but provided no incremental value for T-cell epitope discrimination (change in AUROC +0.004, p = 0.46); attention-derived contact-density was not a valid solvent-accessibility proxy. A curated antibody-epitope benchmark (22 clusters, 5 neutralization-supported) was underpowered for a high-confidence B-cell test. We document data-quality and reproducibility pitfalls (length heterogeneity, coordinate mapping, and pseudoreplication) and release the auditable benchmark. These results provide an auditable candidate resource and show that, in the evaluated benchmarks, ESM-2 sequence scores did not improve epitope prioritization beyond alignment-derived conservation.

genomics↗

bioRxiv2026

Cell-surface N-linked glycans manipulation of K562 cells for augmented susceptibility to natural killer cell killing

Glycosylation of proteins is arguably the most diverse post-translational modification that is altered in almost all cancer types, which has been demonstrated to play a crucial role in creating an immunosuppressive microenvironment that promotes immune tolerance and evasion. However, the biosynthesis of N-linked glycan is mediated by a series of enzymatic reactions catalysed by glycosyltransferases and glycosidases in a template-independent manner, hindering our understanding of specific structure-function relationships and roles of specific glycans on specific proteins. Here we use HLA class I-negative cell line K562, a known reference target for NK-mediated cytolysis, to establish a model investigating how cell surface glycan dynamics influence its susceptibility to cytolysis mediated by NK-92 cells. Treatment of K562 cells with kifunensine, swainsonine, 2F-peracetyl-fucose, or 3Fax-peracetyl Neu5Ac, inhibitors of N-linked glycan processing, resulted in drastic alterations in cell surface carbohydrate phenotype, as could be shown by flow cytometric analysis of the lectiNbinding properties of the cells. Despite these clear changes in carbohydrate phenotype, only K562 cells treated with either kifunensine or 3Fax-peracetyl Neu5Ac exhibited higher susceptibility to the cytolysis medidated by NK-92 cells accompanied with an increased CD107a expression by NK-92 cells. Although K562 cells overexpressing gene MGAT3 exhibited a decreased NK-susceptibility, we further found that this decrease was not exclusively determined by the overexpression of gene MGAT3 product bisecting {beta}1,4-GlcNAc, because the treatment of 3Fax-peracetyl Neu5Ac reversed the resistance of K562 cell against NK-92 cell in despite of expressing higher levels of bisecting {beta}1,4-GlcNAc. Expressing HLA-G on cell surface as extravillous trophoblast did not change the NK-susceptibility of K562 cells, despite evidence that HLA-G molecules expressed by K562 cells can bind to inhibitory receptor ILT2 expressed on NK-92 cell surface. These findings suggest that the level of terminal sialylation, outweighing other components in N-linked glycan, determines the NK-susceptibility of K562 cell, offering a new strategy to weaken the resistance of cancer cells so that the immune system can maximise the elimination.

immunology↗

bioRxiv2026

Niche-specific microbial community structure of subgingival plaque in periodontitis

Subgingival biofilms in periodontitis exhibit spatial heterogeneity, yet the organization of microbial communities across periodontal niches remains incompletely defined. Using paired sampling and 16S rRNA gene sequencing, we characterized non-attached and attached subgingival plaque from patients with periodontitis, together with non-attached plaque from periodontally healthy individuals. Across diversity metrics and ordination analyses, non-attached plaque from periodontitis patients occupied positions between healthy-associated and attached-plaque communities. Taxonomically, these communities contained both health-associated commensals and anaerobic genera commonly enriched in periodontitis. Network analysis identified differences in association-network topology among niches, with the non-attached periodontitis network containing more retained associations than the healthy network. These cross-sectional results describe niche-associated patterns of subgingival community composition and association structure. They do not establish temporal progression, direct microbial interactions, or clinical utility.

microbiology↗

bioRxiv2026

Ecological trajectories and microbial network reorganization across caries-associated oral niches

Dental caries is a biofilm-mediated disease associated with ecological changes in the oral microbiome. How microbial community organization differs among healthy plaque, caries-associated plaque, and carious dentin remains incompletely defined. We used 16S rRNA gene sequencing to profile paired supragingival plaque and carious dentin samples from patients with caries, together with supragingival plaque from healthy controls. Caries-associated plaque showed higher diversity than healthy plaque, whereas diversity was lower in carious dentin. Ecological ordering placed the three sample types along a health-plaque-dentin continuum. Association-network analysis showed distinct network structures in caries-associated plaque and carious dentin, with the dentin network displaying greater density and lower modularity. By integrating differential-abundance and network centrality results, we identified taxa associated with the dentin niche. A sparse logistic regression model using three genera distinguished plaque from dentin in patient-grouped cross-validation (AUROC, 0.780; AUPRC, 0.718). These cross-sectional findings describe niche-associated microbiome organization in dental caries and provide candidate features for future validation in independent, clinically relevant cohorts.

microbiology↗

bioRxiv2026

Evolution-inspired multi-objective Bayesian optimization for protein engineering

Protein engineering requires efficient navigation of vast sequence spaces under limited evaluation budgets, especially when multiple properties must be optimized simultaneously. We developed Evolution-inspired Multi-Objective Bayesian Optimization (EvoMOBO), an active-learning framework that integrates path-dependent sequence generation, global competition among generated variants, and explicit multi-objective optimization. Benchmarking against state-of-the-art methods on complete steroid receptor DNA-binding domain and ParD3 antitoxin landscapes demonstrated robust target-region enrichment, Pareto-front advancement, and sequence diversity across two- and three-objective tasks. In the DBD landscape, simulation-derived geometric descriptors served as labels for both initialization and iterative updating, enriching variants with favorable measured activities without experimental labels. Building on this validation, we applied EvoMOBO to two enzyme-engineering tasks using simulation-derived mechanistic descriptors, with experiments reserved for final validation. For an old yellow enzyme (GkOYE), 16 of 26 tested variants outperformed the wild type, and the best increased non-native oxidative dehydrogenation conversion from 17.5% to 95%. For a formate oxidase (AoFOx), EvoMOBO identified aggregation-resistant variants, two of which nearly doubled diethyl phthalate degradation in a photoenzymatic cascade. Together, these results establish EvoMOBO as a modular framework for multi-objective protein engineering using experimental or mechanism-derived labels.

bioengineering↗