Biology subjects

Shao, Y.

Publications and source records attributed to Shao, Y..

At least 19 recordsLinked to original sources

bioRxiv2026

Dairy cattle movement networks govern the spread and persistence of H5N1 B3.13 in the United States

Avian influenza virus H5N1 is predominantly found in wild and domestic birds. H5N1 genotype B3.13 was first detected in United States dairy cattle in March 2024 and spread through much of the country during 2024 and 2025, resulting in the largest known H5N1 outbreak in domestic mammals. By analysing viral genomic data jointly with historical cattle movement patterns, we find that live cattle movement is the predominant contributor to H5N1 spatial spread, rather than wild or domestic birds. Using phylogeographic methods we estimate that >40% of interstate viral transmissions occurred before B3.13's first detection and >70% occurred before interstate dairy cattle movement restrictions were implemented in April 2024. After this implementation, interstate transmission slowed but within-state spread persisted, driven by local cattle network density and movement volume. Our epidemic simulations demonstrate that the connectivity of counties within and around the Texas Panhandle - where the earliest cases were documented - predisposed the region to become an outbreak epicentre. Counterfactual simulations indicate that interstate restrictions needed to have begun 2-3 months earlier in order to limit the epizootic substantively, given the extent of cryptic spread before detection. Network-informed interventions outperformed geographically-defined containment, but response speed (rather than intervention strength) was the limiting factor in our epidemic models, underscoring the need for surveillance systems capable of detecting pathogen emergence before widespread dissemination.

evolutionary biology↗

bioRxiv2026

A polysynaptic hypothalamuspelvic-ganglion circuit regulates prostate tumor growth via WNT3A

Brain-tumor communication is increasingly recognized. Peripheral autonomic nerves regulate prostate cancer progression, yet how the brain influences local nerve-tumor interactions remains unclear. We identify a polysynaptic circuit linking the hypothalamic paraventricular nucleus (PVN), dorsal motor nucleus of the vagus (DMV), major pelvic ganglion (MPG) and prostate tumor in mice. Although both nuclei were activated during tumor growth, PVN activity restrained, whereas DMV activity promoted tumor growth. MPG ablation abolished DMV-induced tumor promotion. Combining LMD with Smart-seq2 revealed transcriptional remodeling and increased Wnt3a expression in tumor-innervating, activated MPG neurons under upstream neural control. MPG explants released WNT3A, whose neutralization suppressed explant-induced cancer-cell proliferation. Together, our findings identify a polysynaptic PVN-DMV-MPG circuit that regulates prostate tumor growth and implicate WNT3A as a mediator of the pro-tumor effects of the MPG, highlighting the therapeutic potential of targeting the neural circuit and downstream molecular pathway.

cancer biology↗

bioRxiv2026

Programmable design of functional RNAs through multimodal generative learning

RNA engineering is bottlenecked by the tight coupling among sequence, structure, and function, leaving the field reliant on heuristics or screening. We introduce RNAverse, a multimodal generative model that learns a conditionable design grammar from RNA sequences, experimental structural probing profiles, and functional annotations. Beyond setting a new state of the art across 16 predictive modeling tasks, RNAverse inverts into generative design across four distinct RNA classes through three regimes, requiring experimental testing of no more than 20 candidates per-task. Zero-shot in-painting generates snoRNAs that promote premature termination codon (PTC) readthrough, and chimeric aptazymes for ligand-controlled gene regulation, exceeding rational design or large-scale screens. In both settings, RNAverse outperforms family-adapted sequence-only baselines. Fine-tuned in-painting extends to an unseen functional class, yielding IGFBP3-binding aptamers with picomolar affinity that outperform SELEX winners. Iterative optimization engineers 5' UTRs that outperform clinical benchmarks, enhancing protein production in cell-based assays for both the Zolgensma gene therapy and the BNT162b2 COVID-19 vaccine, and boosting antibody responses to BNT162b2 in mice. Together, these results establish multimodal generative learning as a route from predictive RNA modeling to programmable specification of functional RNAs.

bioinformatics↗

bioRxiv2026

Pop-Corn: Predicting Perturbation Phenotype Effects Across Single-Cell and Spatial Contexts

Genetic perturbations can reshape cell populations by altering the relative abundance of specific cell types and states within the profiled population, including increases, decreases and states that become detectable after perturbation. Pooled single-cell screens, such as Perturb-seq, measure such responses at scale. However, only a small fraction of possible perturbations can be tested experimentally. A central challenge is therefore to predict compositional shifts induced by unseen perturbations. Many perturbation-prediction methods do not directly optimize for this outcome; instead, they predict gene-expression responses and infer cell-type and cell-state composition downstream. Surprisingly, we find that even models that accurately predict perturbation-induced changes in average gene expression perform poorly at forecasting these compositional shifts. To address this gap, we present Pop-Corn, a method that directly predicts how a perturbation reshapes cell-type composition without reconstructing gene expression. In the primary T-cell benchmark, Pop-Corn predicted the overall cell-state composition of held-out perturbations more accurately than the evaluated expression-prediction pipelines, while better preserving the diversity of observed cell states. We further extend Pop-Corn to intact tissue, where it predicts perturbation-induced cell-type proportion changes in local cellular neighborhoods and uses attention patterns to generate hypotheses about context-dependent cellular interactions. Retrospective virtual screens support the use of Pop-Corn to prioritize perturbations for experimental follow-up according to their predicted effects on cell-state composition.

bioinformatics↗

bioRxiv2026

From pelagic to reef: Genomic basis of morphological adaption during life-history transition in the orbicular batfish (Platax orbicularis)

Metamorphosis enables teleosts to transition between ecological niches, driving lineage-specific morphological innovations. The orbicular batfish (Platax orbicularis) exemplifies this, transforming from a pelagic larva into a reef-associated juvenile with extreme allometric fin elongation. Here, we generated a chromosome-level genome assembly of 715.46 Mb, with 99.4% completeness. Integrating comparative genomics with tissue-specific transcriptomics, we demonstrate that such transformation is a systemic process shaped by both long-term evolutionary adaptation and short-term developmental reprogramming. Comparative analyses revealed expansion of development-related gene families and significant evolution of immune- and sensory-related families, likely associated with ecological niche transition. Spatiotemporal clustering of metamorphosing fin transcriptomes identified rapid cell cycle activation accompanied by sustained suppression of basal metabolism, indicating developmental trade-offs during fin elongation. Several differentially expressed genes involved in cell cycle (ankfn1 and mcph1), ossification (cyp24a1 and sema4d), and extracellular matrix remodeling (sdc4 and serpine1) exhibited positive selection, suggesting coordinated regulatory and protein-level adaptation. Among an extensive repertoire of 49 hox genes, four (hoxb6b, hoxc8a, hoxd9b, hoxd10a) were differentially expressed, implicating developmental regulators in fin remodeling. Together, our findings offer new insights into the evolutionary context and molecular mechanisms of metamorphosis in orbicular batfish, and advance the understanding of adaptive evolution and metamorphosis innovations in teleosts.

genomics↗

bioRxiv2026

A multimodal representation learning platform for accurate molecular ADMET prediction

Accurate ADMET prediction is essential for prioritizing compounds before costly experimental validation, yet ADMET tasks are highly heterogeneous. Properties such as solubility, permeability, protein binding, clearance, transporter activity and toxicity are governed by different molecular signals, ranging from local functional groups and physicochemical descriptors to bonded topology and three-dimensional geometry. Consequently, a single molecular representation or backbone is unlikely to be optimal across all ADMET tasks. We present Trimole-Hybrid, a task-wise multimodal framework that addresses ADMET heterogeneity by selecting or combining predictors built from complementary molecular representations. Trimole-Hybrid constructs a candidate pool of SMILES-, graph-, geometry-sensitive EPT/3D- and chemical descriptor-based predictors. For each task, Trimole-Hybrid selects the best-performing predictor to obtain the final prediction. On 22 Therapeutics Data Commons ADMET benchmarks, Trimole-Hybrid exceeded the public TDC top-1 methods on 10 tasks and ranked within the top 10 for 21 tasks. Ablation studies confirmed the contribution of both complementary multimodal molecular representations and task-specific ensemble strategies. In two small-molecule case studies, Trimole-Hybrid shows sensitivity to changes in essential functional motifs, suggesting its ability to capture ADMET-relevant molecular substructures.

bioinformatics↗

bioRxiv2026

Quantitative Modeling of TLR Signaling Reveals Missing Negative Feedback Guiding Identification of TANK-IKKε Checkpoint

Toll-like receptor (TLR) signaling must be activated rapidly and then terminated to support host defense without sustained inflammation. We developed a rule-based model of mouse macrophage TLR4 signaling at the molecular-interaction level using measured protein copy numbers, RNA-seq-based abundance estimates, literature- and structure-informed reaction rates, and 979 dynamic experimental constraints. The trained model reproduced much of the TLR4-induced NF-{kappa}B and MAP kinase response but consistently failed to capture deactivation of MyD88, TRAF6-associated species, and IKK/{beta}. The recurrent model failure conveyed important biological information, localizing missing regulation to the proximal MyD88-IRAK-TRAF6 module and guiding experimental evaluation of IKK{varepsilon} and its scaffold TANK. Loss of IKK{varepsilon} enhanced transcriptional, cytokine, MAP kinase, and NF-{kappa}B responses to MyD88-specific TLR ligands. TANK deficiency produced a similar cellular phenotype and abolished stimulus-induced IKK{varepsilon} phosphorylation. Deficiency of either protein increased IRAK1 and TRAF6 ubiquitination without increasing MyD88 ubiquitination, placing the inhibitory checkpoint at or immediately downstream of the IRAK1-TRAF6 ubiquitin-signaling node. Overlapping but non-identical in vivo phenotypes further supported a shared regulatory axis with additional protein-specific functions. Our study presents a model-experiment discovery cycle where quantitative pathway discordance identifies missing biology and reveals a TANK-dependent IKK{varepsilon} checkpoint that restrains MyD88-driven inflammation.

systems biology↗

bioRxiv2026

MxSure: a mixture model for inferring within-host substitution rates and transmission SNP thresholds

Quantifying short-term evolutionary rates of microbial genomes is essential for understanding the processes that shape within-host evolution and for establishing thresholds needed to track transmission. In studies of short-term evolutionary rates, samples are often collected from closely related clusters (e.g. longitudinally from the same host or from transmission pairs), with substantial time intervals separating genomes between clusters. Distinguishing strain replacement from persistence presents is also difficult in these studies. In addition, many public health and metagenomic bacterial strain tracking pipelines output pairwise SNP distances rather than the multiple sequence alignments required by common substitution rate estimation pipelines. This makes it hard to estimate within-host evolutionary rates in many commensal bacterial species that are difficult to culture and isolate. To address these challenges, we introduce MxSure, a tool for estimating substitution rates and transmission thresholds while accounting for strain replacement from pairwise SNP distance data, as commonly generated by transmission tracking and metagenomic analysis pipelines. We demonstrate the accuracy of MxSure through extensive simulations and by analysing species with previously estimated substitution rates from longitudinal metagenomic datasets. Using MxSure, we estimated within-host substitution rates and transmission SNP thresholds for multiple commensal bacterial species including Bifidobacterium longum and Bifidobacterium bifidum from a longitudinal study of the infant gut microbiome.

bioinformatics↗

bioRxiv2026

Real Science Is Harder Than Benchmarks: Evaluating Advanced AI Frameworks on Published Studies. I. Uncertainty Quantification, ML on Therapeutic Data Commons, and Agent-Based Modeling

Artificial Intelligence (AI) frameworks for automating scientific research have shown strong performance on benchmarks, but their capacity to routinely reproduce results from multiple real-life published studies remains largely untested. We evaluated five advanced AI research frameworks (Kosmos, K-Dense, ToolUniverse, BioAgents from bio.xyz, and the AI Scientist-v2 from Sakana AI) on three real-life tasks (including two recently published papers) spanning uncertainty quantification for molecular property predictions, machine learning on Therapeutic Data Commons benchmarks, and agent-based modeling. AI frameworks demonstrated genuine strengths: generating original hypotheses, competently executing routine data acquisition and coding tasks, providing statistical measures of confidence often absent from the original papers, and producing well-formatted final reports. At the same time, our experiments revealed that real-world scientific tasks remain considerably harder than current benchmarks suggest. No AI framework matched the scope or depth of the original studies, results varied across multiple runs of the same framework with the same prompt, and we documented cases of severe hallucinations in final reports, gaps in literature coverage, and overconfident conclusions. Verification of AI outputs required substantial domain expertise. While these three tasks are only partially representative of the broader scientific landscape, they offer a starting point for developing a more rigorous methodology for evaluation of AI performance than what is currently practiced. We conclude that AI frameworks are already valuable for prototyping research directions and stress-testing completed studies, and some of the limitations documented here appear largely tractable through infrastructure improvements and continued development.

bioinformatics↗

bioRxiv2026

Reactivation of a TAL1 progenitor cell enhancer region by non-coding somatic variants in T-lineage acute lymphoblastic leukemia

Aberrant activation of TAL1, a key oncogenic driver, defines a major subgroup comprising [~]30% of childhood T-lineage acute lymphoblastic leukemias (T-ALLs). We and others have shown that somatic non-coding mutations within upstream and intronic cis-regulatory regions of TAL1 contribute to transformation by creating binding sites for MYB and other transcription factors. Here we investigated cis-regulatory mechanisms mediated by somatic mutations occurring in an intergenic region located 29 kilobase pairs downstream of the canonical TAL1 transcription initiation site, implicated in 6% of TAL1-expressing T-ALLs. These somatic variants include i) complex indels resulting in de novo MYB transcription factor binding sites (TFBSs) and ii) internal tandem duplications (ITDs) encompassing canonical MYB TFBSs. Chromatin immunoprecipitation sequencing (ChIP-seq) revealed binding of the TAL1 core regulatory circuit (CRC) transcription factors MYB, GATA3, and RUNX1, resulting in enhancer activity mediated by sequences with the mutant allele. Strikingly, ChIP-seq peaks for the repressive H3K27me3 mark and the active H3K27ac mark co-existed across TAL1 regulatory sequences but enriched for different haplotypes. TAL1 transcription from the mutant haplotype initiated from a promoter located within exon 4 of the canonical TAL1 transcript, resulting in a short isoform normally expressed by hematopoietic stem cells (HSC). Interestingly, neither the isoform expression nor the enhancer activity could be predicted by the sequence-to-function deep learning artificial intelligence (AI) model AlphaGenome, emphasizing the importance of experimental validation. Our findings indicate that selection for cis-regulatory, non-coding variants leads to reactivation of enhancers normally active in HSC but silenced in differentiated lineages during normal hematopoietic cell development.

cancer biology↗

bioRxiv2026

Revisiting the Monascus genus (Eurotiales, Aspergillaceae): A Multilocus Phylogenetic Approach to Species Delimitation

Monascus spp. are economically important filamentous fungi that have been utilized in the production of beneficial metabolites such as Monascus pigments and monacolin K, as well as in the brewing of some Asian fermented foods. The delimitation of Monascus species has traditionally relied on phenotypic traits; however, this morphological classification approach is susceptible to subjective judgments and variations in cultural conditions and also may not necessarily be related to the actual genetic relationship. Consequently, synonymy and misidentification frequently occur in Monascus taxonomy, highlighting the urgent need for a convenient and reliable classification system for this genus. In this study, a phylogenetic analysis of 82 representative Monascus strains, encompassing all previously recognized species of the genus, was conducted based on the concordance of five gene genealogies (BenA, CaM, ITS, LSU, and RPB2) to clarify species delimitation and resolve phylogenetic relationships within Monascus. The results revealed that the genus Monascus is resolved into 11 species, which are clustered into two sections: Floridani (including M. argentinensis, M. flavipigmentosus, M. floridanus, M. lunisporas, M. mellicola, M. pallens, and M. recifensis) and Rubri (including M. pilosus, M. purpureus, M. ruber, and M. sanguineus). M. pilosus and M. sanguineus were reaffirmed as distinct species due to their well-supported and divergent phylogenetic lineages. Additionally, M. albidulus, M. anka, M. barkeri, and M. fumeus are synonymized with M. pilosus, while M. aurantiacus and M. rutilus are synonyms of M. purpureus. Finally, a comprehensive list of accepted Monascus species along with their corresponding barcode sequence data is provided.

microbiology↗

bioRxiv2026

IMMREP25: Unseen Peptides

T cell receptors (TCRs) can bind to peptides presented by MHC molecules (pMHC) as a first step to trigger a T cell response. Reliable approaches to predict TCR:pMHC binding would have broad applications in clinical diagnostics, therapeutics, and the fundamental understanding of molecular interactions. IMMREP is a community organized series of prediction contests that asks participants to predict TCR:pMHC binding on unpublished datasets. Previous iterations in 2022 and 2023 showed multiple approaches can predict TCR-pMHC binding with significant accuracy (median AUC_0.1[≥]0.7) for peptides where experimental data is available ("seen" peptides). In contrast, models did not outperform random guessing for peptides that have no such data available ("unseen" peptides). Here we report on the results of IMMREP25, which focused solely on unseen peptides in order to evaluate the cutting edge of the field. We received 126 named submissions predicting the specificity of 1,000 TCRs against twenty unseen peptides restricted by one of two MHC molecules (HLA-A*02:01 and HLA-B*40:01). The best performing methods showed a macro-AUC_0.1 of 0.60, significantly better than random, demonstrating significant advances in the field. The top performing methods incorporated structural modeling into their approach, indicating that especially for unseen peptides, a structural understanding aids in the prediction of TCR:pMHC interactions. The results from this benchmark highlight the significant challenges remaining for TCR:pMHC predictions and will inform future method development.

bioinformatics↗

bioRxiv2026

Dispersal, adaptation and persistence of H5N1 in the sub-Antarctic and Antarctica

High pathogenicity avian influenza virus (HPAIV) H5N1 reached the sub-Antarctic and Antarctica in 2023, subsequently spreading to remote locations within this region where it had devastating impacts on seal, penguin and albatross populations. The threat to marine wildlife over this broad area exemplifies the need to understand H5N1 long-distance dispersal and evolution. We obtained 104 novel viral genomic sequences from samples that we collected at South Georgia, Kerguelen, Crozet, Prince Edward, Falklands/Malvinas Islands and the Antarctic Peninsula in a region spanning 8,000 kilometers. Using recent phylogeographic modeling advances we show that H5N1 spread encompassed numerous transmission events between distant locations, accumulating mammalian-adaptive mutations in the process. Seals are the most affected species, but we reveal that the long-distance eastward virus dispersal better aligns with the long-distance movements of large petrels and albatrosses. The risk of H5N1 endemisation, dispersal to other locations and ongoing evolution are highly concerning.

microbiology↗

bioRxiv2026

A New Information Theoretic Approach Shows that Mixture Models Outperform Partitioned Models for Phylogenetic Analyses of Amino Acid Data

Partitioned and mixture models are widely employed in Maximum Likelihood phylogenetic analyses of large genomic datasets. Comparing the fit of the two types of models has been challenging, because standard information-theoretic approaches cannot be applied. Mixture models are increasingly popular for the analysis of amino acid datasets and can lead to different conclusions compared to partitioned models. This raises an important question - which type of model tends to perform better? Susko et al. (2026) recently introduced the marginal Akaike information criterion (mAIC), which allows mixture models and partitioned models to be directly compared for the first time. Here, we use the mAIC and a range of other approaches to compare the fit of mixture and partitioned models across a diverse set of empirical datasets. We show that mixture models are universally favoured on amino acid datasets. This has important implications for interpreting empirical analyses and suggests that continued development of mixture models is an important avenue for future research.

evolutionary biology↗

bioRxiv2026

Integrative Multi-Scale Sequence-Structure Modeling for Antimicrobial Peptide Prediction and Design

Antimicrobial resistance (AMR) is accelerating worldwide, undermining frontline antibiotics and making the need for novel agents more urgent than ever. Antimicrobial peptides (AMPs) are promising therapeutics against multidrug-resistant pathogens, as they are less prone to inducing resistance. However, current AMP prediction approaches often treat sequence and structure in isolation and at a single scale, leading to mediocre performance. Here, we propose MultiAMP, a framework that integrates multi-level information for predicting AMPs. The model captures evolutionary and contextual information from sequences alongside global and fine-grained information from structures, synergistically combining these features to enhance predictive power. MultiAMP achieves state-of-the-art performance, outperforming existing methods by over 10% in MCC when identifying distant AMPs sharing less than 40% sequence identity with known AMPs. To discover novel AMPs, we applied MultiAMP to marine organism data, discovering 484 high-confidence peptides with sequences that are highly divergent from known AMPs. Notably, MultiAMP accurately recognizes various structural types of peptides. In addition, our approach reveals functional patterns of AMPs, providing interpretable insights into their mechanisms. Building on these findings, we employed a gradient-based strategy and achieved the design of AMPs with specific motifs. We believe MultiAMP empowers both the rational discovery and mechanistic understanding of AMPs, facilitating future experimental validation and therapeutic design. The codebase is available at https://github.com/jiayili11/multi-amp.

bioinformatics↗

bioRxiv2026

Foundation Models Improve Perturbation Response Prediction

Predicting cellular responses to genetic or chemical perturbations has been a long-standing goal in biology. Recent applications of foundation models to this task have yielded contradictory results regarding their superiority over simple baselines. We conducted an extensive analysis of over 600 different models across various prediction tasks and evaluation metrics, demonstrating that while some foundation models fail to outperform simple baselines, others significantly improve predictions for both genetic and chemical perturbations. Furthermore, we developed and evaluated methods for integrating multiple foundation models for perturbation prediction. Our results show that with sufficient data, these models approach fundamental performance limits, confirming that foundation models can improve cellular response simulations. Code and Data: https://github.com/genbio-ai/foundation-models-perturbation

bioinformatics↗

bioRxiv2026

Modeling mitochondrial inheritance enables high-precision single-cell lineage tracing in humans

Somatic mutations in mitochondrial DNA (mtDNA) provide natural barcodes that enable engineering-free lineage tracing in human tissues, but the complex dynamics of mtDNA inheritance across cell divisions and incomplete sampling of mtDNA introduce uncertainty in reconstructed lineages. Here, we present MitoDrift, a probabilistic framework that integrates Wright-Fisher drift dynamics with sparse single-cell measurements to produce confidence-refined lineage trees enriched for accurate clonal relationships. Validation with gold-standard lentiviral barcoding and whole-genome sequencing demonstrates that MitoDrift outperforms existing tree reconstruction methods in precision while maintaining high clonal recovery, enabling robust analyses linking lineage to cell state. Applying MitoDrift to human hematopoiesis reveals an age-associated decline in clonal diversity with differential impact across cell types and identifies heritable regulatory programs in hematopoietic stem cells in vivo, linking AP-1/stress-associated programs to clonal expansions. In multiple myeloma, MitoDrift captures therapy-associated clonal remodeling undetectable by copy number analysis, revealing phenotypic transitions and linking gene regulatory programs to differential drug sensitivity. Collectively, MitoDrift enables high-precision lineage tracing at scale and establishes quantitative lineage-state analysis in primary human tissues, linking clonal history to transcriptional and epigenetic programs in tissue homeostasis, aging, and disease.

genomics↗

bioRxiv2026

Benchmarking generative scaffold design methods for peptide engineering in TCR-MHC complexes

De novo peptide design at T cell receptor-peptide-major histocompatibility complex (TCR-pMHC) interfaces is a central challenge in computational immunology, with direct implications for vaccine development, cancer immunotherapy, and autoimmune disease. Despite rapid advances in generative protein modeling, there is currently no systematic benchmark evaluating these methods in the highly constrained and immunologically relevant setting of peptide-MHC presentation and TCR recognition. Here, we present two complementary contributions. First, we introduce a multi-stage computational pipeline for peptide design in predefined TCR-pMHC contexts, integrating generative modeling with sequence optimization and structure-based filtering. Second, we establish a benchmark for evaluating generative peptide design methods in TCR-pMHC complexes. Using a curated dataset of high-quality crystal structures deposited after the AlphaFold3 training cutoff, we assess state-of-the-art generative approaches for peptide backbone generation, sequence design, and the enrichment of near-native solutions. We explicitly examine whether different backbone generation strategies respect the geometric constraints of the MHC binding groove and recover native-like peptide conformations. Our results reveal substantial method-dependent differences: some generative strategies fail systematically in the groove-bound peptide setting, whereas others generate physically plausible backbones with varying accuracy and conformational diversity. We further show that enforcing anchor constraints strongly influences peptide conformations at non-anchor positions, highlighting a trade-off between structural accuracy and conformational sampling. To enable fair and reproducible comparison, we introduce a standardized, multi-stage scoring protocol that integrates MHC binding prediction, physics-based energy evaluation, and independent structure prediction confidence metrics to enrich near-native designs from large candidate pools. Together, this work establishes the first comprehensive pipeline and benchmark for generative peptide design at TCR-pMHC interfaces and provides practical guidelines for developing peptide design workflows and evaluating generative models in immunologically constrained protein design settings.

bioinformatics↗