ImmunoInformatics

Predicting linear B-cell epitopes is a basic immunoinformatics task that has a direct impact on vaccine design and antibody engineering. Recent advances in machine learning have improved predictive performance, but most existing approaches are trained on aggregated datasets and assume that antigenic patterns are conserved across host organisms. This assumption ignores the immunological variability depending on the host and prevents generalizing the model across species. This is the first systematic host-wise evaluation where we present a systematic machine learning-based analysis of host-aware linear B-cell epitope prediction using curated datasets from the Immune Epitope Database (IEDB). We build separate datasets for human, mouse, and non-human primate hosts and assess several classification models, including Random Forest, Support Vector Machine (SVM), Gradient Boosting, XGBoost, and K-Nearest Neighbors (KNN). The models exploit feature representations derived from sequences, such as AAIndex descriptors, biochemical properties from ExPASy, and dipeptide composition. Our results show that predictive performance differs substantially across hosts. Models achieve up to 86.07% accuracy and 0.93 ROC-AUC on human datasets but lower performance on mouse and non-human primate datasets. This gap underlies dataset bias and sequence distribution differences, as well as the inability of existing features to capture host-specific immunological context. These results indicate that the prediction of linear B-cell epitopes is intrinsically host-specific, and a single global model does not generalize well across species. We propose to incorporate host-aware modeling strategies and organism-specific features for enhanced predictive reliability and biological relevance.

Determining the structural basis of antigen recognition by antibodies and T cell receptors (TCRs) provides critical insights into effective immune targeting and can inform design of biotherapeutics and vaccines. Accurate computational modeling of antibodies and TCRs in complex with their targets poses a major challenge for predictive methods, including AlphaFold, which is generally accurate for modeling protein complexes but has shown limited success for immune recognition. In this study we assessed the performance of AlphaFold2, AlphaFold3, increased sampling protocols, and related deep learning methods for modeling antibody-protein, antibody-peptide, and TCR-peptide-major histocompatibility complex (pMHC) recognition. We show that increased sampling and AlphaFold3 generally improve performance relative to default sampling and AlphaFold2, however predictive accuracy and improvement levels varied considerably among interface classes, with antibody-peptide complexes representing a challenge despite their small antigen size. Comparing per-case success across methods showed some complementarity, indicating opportunities for increased success through model pooling approaches, for instance increasing antibody-peptide near-native success from 41% to 59%. Analysis of AlphaFold confidence scores and modeling of a noncanonical complex provided further insights into predictive performance. These results highlight considerations for predictive antibody and TCR complex modeling efforts, while revealing key distinctions among protocols, scoring, and immune complex classes.

Prediction of B-cell epitopes can assist in reducing costly wet-lab screening in vaccine design, diagnostics, and antibody discovery. However, current predictors often suffer from noisy labels, weak generalization, and structure-dependent workflows. Here we present EO_SCPLOWPIC_SCPLOWESM-GA, an efficient sequenceonly pipeline for linear B-cell epitope prediction. Positive and negative peptide examples are collected from IEDB, which provides experimentally tested epitopes and distinguishes positive and negative epitope records based on assay evidence(Vita et al., 2019). Each peptide is encoded with a frozen ESM-2 protein language model: a bidirectional transformer producing amino acid embeddings for downstream structure and function tasks (Lin et al., 2023). Mean-pooled embeddings are further compressed into a compact 420-feature representation with a genetic algorithm and classified with lightweight Random Forest, XGBoost, or MLP heads. This avoids foundation-model fine-tuning, reduces the number of trainable parameters, improves interpretability, and enables low-resource deployment. On an IEDB-derived benchmark, EO_SCPLOWPIC_SCPLOWESM-GA attains 0.880{+/-} 0.004 AUC-ROC, 0.852{+/-} 0.005 PR-AUC, 82.0 {+/-} 0.6% accuracy, 0.79 {+/-} 0.01 F1, and 0.74{+/-} 0.01 MCC, outperforming dense ESM-2 features and baselines LBCE-XGB, EpitopeVec, and BepiPred-2.0 (mean{+/-} std over five independent random seeds). The framework shows how frozen protein foundation models can enable pandemic preparedness, peptide vaccine prioritization, diagnostic antigen screening, and equitable computational immunology.

Accurate immune receptor design requires modeling the coupled variation of amino-acid sequence, full-atom conformation, and target-binding geometry across antibodies, nanobodies, and T-cell receptors (TCRs). Existing methods often address only part of this problem, either by separating structure generation from sequence design, relying on fixed-backbone inverse folding, or focusing on a single receptor class. We introduce IgGM2, a unified all-atom generative framework for immune receptor structure prediction and CDR sequence-structure co-design. IgGM2 follows a structure-to-design strategy: it first learns how immune receptors are positioned around fixed target structures, and then transfers this target-conditioned structural prior to CDR design. Unlike modular design pipelines, IgGM2 jointly generates CDR residue identities and full-atom receptor structures, allowing framework geometry to adapt to designed CDRs without separate inverse folding or external sidechain packing. Unlike continuous residue encodings based on virtual-atom geometry, IgGM2 keeps sequence prediction explicit while using atom14 placeholders only for full-atom representation. On structure prediction benchmarks, IgGM2 better captures receptor-target spatial relationships than AlphaFold3 on FoldBench and achieves strong performance on TCR-pMHC modeling. On sequence design benchmarks, IgGM2 improves amino-acid recovery and Rosetta-based interface preference metrics, suggesting more favorable generated binding interfaces. These results support IgGM2 as a unified all-atom framework for adaptive immune receptor structure prediction and design.

Predicting how mutations alter antibody-antigen binding affinity is essential for antibody engineering and vaccine design, yet current methods generalize poorly to unseen complexes. We present a multi-scale machine learning framework integrating 93 descriptors across four modalities: physicochemical, structural, ESM-2 protein language model, and solvent-accessible surface area (SASA)/{Delta}{Delta}Gfold features. Under leave-one-complex-out deep mutational scanning (LOCO-DMS) cross-validation on AbAgym (36,541 mutations, 68 experiments, 13 pathogens), gradient boosting achieved MCC = 0.206; a confidence-stratified ensemble reached MCC = 0.374 (83.5% accuracy, 25.5% coverage). No single modality exceeds the majority baseline alone; only multi-scale fusion succeeds. Boltzmann ceiling analysis shows 45.9% of mutations are near-neutral (|{Delta}{Delta}G| < kBT), bounding theoretical maximum MCC at 0.473; our method achieves 79.1% of this limit. Five deep learning architectures benchmarked under LOCO-DMS showed self-attention matching gradient boosting (MCC = 0.200). Cross-pathogen transfer failed systematically (mean 46.7%), confirming universal binding predictors remain an open challenge.

Germinal is a recently described computational pipeline for de novo antibody design that combines AlphaFold-Multimer hallucination with antibody language model guidance to generate epitope-targeted antibodies. Germinal identified binders with nanomolar-to-low-micromolar affinities by testing only 43-101 designs per target across four diverse antigens, establishing it as a practical tool for epitope-directed antibody design accessible to standard academic laboratories. As this architecture is itself very recent, systematic replacement and benchmarking of its individual components remains largely unexplored, yet offers a valuable opportunity to probe the robustness of the underlying design. We present OpenGerminal, which replaces PyRosetta with a fully open-source stack comprising OpenMM 8.5.1, FreeSASA, FASPR, Biopython, and sc-rs v1.0.0, and adopts AbLang1 (ablang2 v0.2.1) as the sole antibody language model in place of IgLM. Benchmarking on two VHH targets (PD-L1 and IL-3) reveals that OpenGerminal achieves a markedly higher cofolding pass rate (PD-L1: 33.7% vs. 18.6%; IL-3: 24.6% vs. 8.0%) with equivalent or improved Chai-1 structural confidence metrics in accepted designs, at the cost of a modest increase in per-trajectory computation time (>=1.5x). Multi-chain target support is also extended and verified to run without error on the official insulin example. OpenGerminal provides the first systematic benchmarking of IgLM versus AbLang1 within the Germinal architecture, and its fully open-source component stack broadens the range of deployment contexts in which the pipeline can be used.

Influenza A viruses H3N2 and H10N5 represent, respectively, a persistently dominant seasonal pathogen and a newly documented zoonotic threat with the latter strain variants responsible for the first confirmed human fatality in January 2024, yet no vaccine platform currently addresses co-protection against both subtypes within a unified immunogen. We report here the immunoinformatics based vaccine design and multi-layered computational validation of a 419-amino-acid multi-epitope subunit vaccine construct targeting conserved hemagglutinin (HA) and neuraminidase (NA) antigens identified through multiple sequence alignment of the avian H10N5 (A/swine/Hubei/10/2008) and H3N2 human reference strain sequences to identify viral agents undergoing mammalian adaptations. Linear B-cell, cytotoxic T lymphocyte (CTL), and helper T lymphocyte (HTL) epitopes were predicted using ABCpred, BCEpred, BepiPred 2.0, NetMHCpan 2.1, and NetMHCpan 4.0, then filtered through VaxiJen 3.0, AllerTOP v2.1, and ToxinPred to retain only antigenic, non-allergenic, non-toxic candidates. The final construct, incorporating an avian {beta}-defensin N-terminal adjuvant with GPGPG, AAY, and EAAAK linkers, exhibited a molecular weight of 43.9 kDa, instability index of 31.15, and SOLPro solubility probability of 0.763. Tertiary structure modeling via I-TASSER and GalaxyRefine achieved 84.4% Ramachandran-favored residues. Molecular docking against TLR3 and TLR7 yielded binding free energies of -16.1 and -16.8 kcal/mol with picomolar dissociation constants. Molecular dynamics simulations confirmed complex stability over extended trajectories. Furthermore, codon optimization produced a Codon Adaptation Index of 1.0 for E. coli K12 expression. In silico immune simulation demonstrated robust activation of humoral and cellular immunity including elevated IgG1, IgM, IFN-{gamma}, IL-2, rapid NK cell expansion, and broad B-cell clonal diversity. These findings establish a computationally validated candidate capable of providing protection against influenza in multiple host organisms, warranting experimental advancement.

BackgroundTuberculosis remains a leading global infectious killer, with BCG offering inconsistent adult protection and rising drug-resistant strains demanding novel vaccine strategies. We report the first multi-epitope vaccine construct simultaneously targeting three previously unexplored Mycobacterium tuberculosis virulence proteins; EccB3, MycP, and polyketide synthase which collectively govern nutrient acquisition, ESX secretion integrity, and innate immune evasion. MethodsUsing a reverse vaccinology pipeline, B-cell, CTL, and HTL epitopes were predicted, filtered for allergenicity, toxicity, and IFN-{gamma} induction, then assembled into an 823-residue chimeric construct incorporating beta-defensin and PADRE adjuvants with AAY/GPGPG linkers, covering [~]90% global HLA diversity. The construct underwent AlphaFold structure prediction, 3DRefine refinement, disulfide engineering, PROCHECK/ProSA validation, ClusPro 2.0 docking against TLR1/TLR2, and C-IMMSIM immune simulation. ResultsThe construct (82.3 kDa, instability index 32.48) showed strong structural quality (94.7% favoured Ramachandran residues), stable TLR1/TLR2 binding (weighted energy: -1,371.0 kcal/mol), and robust in silico immune responses and durable memory cell formation following booster simulation. ConclusionThis computationally validated construct represents a promising multi-target TB vaccine candidate warranting experimental advancement.

BackgroundLate-stage attrition in therapeutic antibody discovery is dominated by developability liabilities: aggregation, polyspecificity, charge-driven non-specific binding, and chain-mispairing artefacts. Bispecific antibodies amplify these risks because each additional binding arm adds a new biophysical envelope that must be jointly satisfied. The existing in-silico ecosystem addresses individual axes of this problem (humanization, structure prediction, single-metric developability scoring) but few platforms integrate them end-to-end. PTIm-mAb (SANSHI Bio Solutions Pvt Ltd) is a multi-objective, AI/ML-driven antibody design platform that jointly optimizes sequence liabilities, surface aggregation, charge balance, humanness, and predicted binding affinity, and recommends a bispecific architecture in a single workflow. MethodsWe applied PTIm-mAb to the published sequences of eleven FDA-approved bispecific antibodies using the platforms default-parameter Pareto-acceptance optimization loop, run to convergence or to the internal iteration ceiling, with no human curation between the platform run and the external profiler. Both wild-type and platform-optimized sequences were profiled independently with three publicly available developability tools: Aggrescan, CamSol, and the Therapeutic Antibody Profiler (TAP). Paired-sample tests (Wilcoxon signed-rank, exact binomial sign test, McNemar exact test) evaluated the direction and significance of changes. ResultsAcross the 17 evaluable paired arms profiled by TAP, PTIm-mAb cleared four wild-type CDR-vicinity Positive Charge Patch (PPC) flags Blinatumomab-Arm1 (1.9952 [-&gt;] 0.6885), Mosunetuzumab-Arm1 (1.3391 [-&gt;] 0.0568), Linvoseltamab-Arm2 (0.8060 [-&gt;] 0.0), and the headline Elranatamab-Arm1 case (1.7981 [-&gt;] 0.5799) achieved without trading off any other in-range metric and corroborated by Aggrescan and CamSol on the same arm. Total CDR length was significantly shortened across the cohort (Wilcoxon two-sided p = 0.0075, one-sided p = 0.0037, effect size r = 0.65): significant improvement on the metric most directly under the optimizers control. The directional shift on Aggrescan integrated aggregation propensity was also significant by sign test (24 of 36 chains improved, 2 unchanged, 10 worsened; p = 0.021). On the already-clean Zenocutuzumab profile the optimizer identified residual headroom (PPC 0.1191 [-&gt;] 0.0; SFvCSP 12.5 [-&gt;] 6.0), demonstrating that the platforms value extends to candidates that pass all flags. Three results: Teclistamab Arm-1, Emicizumab, and Talquetamab Arm-2 did not clear all flags and are presented as candidates for iterative re-invocation of the platform pipeline on the optimized output (planned follow-up; Section 5). The remaining TAP metrics (PSH, PPC magnitude, PNC, |SFvCSP|) trended in the improvement direction without reaching significance in this cohort, a pattern consistent with the expected statistical signature of a multi-objective optimizer applied to molecules already within the clinical-stage envelope. The platform reported a mean of 12.8 months and USD 723,889 of computational front-loading per project across the nine-project cohort (range 9.0-16.0 months; USD 510,000-960,000); the underlying cost assumptions are tabulated in Supplementary Table S3. ConclusionPTIm-mAb produces externally verifiable, literature-aligned improvements on the metrics most directly under its control, clears CDR-vicinity charge-patch flags on a meaningful fraction of flagged candidates, and front-loads substantial design-iteration work. The cohort-level pattern is consistent with a calibrated multi-objective optimizer operating at the edge of detectable headroom on a deliberately hard benchmark. We position the platform as an early-stage triage and lead-optimization layer in bispecific antibody discovery. For molecules whose first-pass result does not clear all flags, iterative re-invocation of the pipeline on the optimized output is a natural follow-up direction.

Reported peptide-MHC (pMHC) AUROCs of 0.85-0.95 overstate generalization to unseen alleles: because immunopeptidome data are dense on a few well-studied alleles and sparse on the rest, training and test sets come to share near-identical alleles, so the numbers partly reflect interpolation rather than extrapolation to new MHC grooves. This is a property of the data, not of any one method. We assembled an open, harmonized corpus of 5.8 million experimental measurements across both HLA classes and use it to control the leakage explicitly: alleles held out at the sequence and cluster level, peptide-disjoint splits, and provenance-matched negatives. On strictly novel alleles, generalization is in the high 0.7s rather than the 0.9s a conventional split returns. Against this benchmark we trained a predictor that spans both classes in one model and factors presentation into a peptide-only ligand-likeness term and an allele-specific term; it exceeds eight published predictors by per-allele {Delta}AUROC = +0.22 to +0.37 (p < 10-9), most on the least-studied genes. Corpus, benchmark, and model are released. Author summaryOur immune cells display protein fragments on the cell surface, held by molecules (the human leukocyte antigens, or HLAs) that vary from person to person. Predicting which fragments a given HLA displays matters for cancer vaccines, transplant matching, and the safety of engineered therapies, and many computational tools now do it well. Most available data come from a few common HLAs, so test cases tend to resemble training cases, and the published accuracy looks better than it really is for the rare HLAs that matter most in the clinic. We assembled a large, openly shared collection of experimental measurements across both major HLA classes and used it to test prediction more directly, holding out HLAs that are sequence-distant from those in training. Accuracy on these is measurable but lower than the usual figures suggest. We also built a predictor that handles both HLA classes in one model and gains most relative to existing tools on the rare HLAs where they are weakest. The data, benchmark, and model are available for the same test.

Structure prediction models have moved from single proteins to assemblies that include diverse biomolecules and their modifications. AlphaFold3 (AF3) and related models extended structural modelling via an all-atom framework, opening many new potential applications in structural biology. We evaluate how well the new capabilities of AF3 translate into application tasks in diverse areas: prediction of ubiquitinated protein structures, T-cell receptor (TCR)-epitope recognition, antibody-antigen complexes, protein-RNA and protein-lipid interactions. We find that, while AF3 can perform well in favourable settings, this performance is uneven across applications. In RNA-target predictions, the model confidence fails to separate genuine from decoy interaction partners and in several tasks accuracy depends on the presence of related complexes in the training set. Taken together, our assessment is more cautious than for AF2, whose gains in modelling monomers and complexes were clear and broadly generalisable. AF3s extension to new biomolecule types shows less consistent performance and generalisation. AF3 can be a powerful tool for hypothesis generation and prioritisation, but its predictions and use of confidence metrics will depend strongly on the specific application area and must be interpreted with respect to training-set overlap. We expect that the benchmarks provided here will serve for testing of future developments in the structure prediction field.

Blood group antigens, defined by epitopes on the erythrocyte surface, are central to transfusion safety and maternal-fetal compatibility. While the genetic basis of many clinically relevant blood group antigens is well established, which structural and biophysical parameters determine whether a single-nucleotide variant gives rise to an antigenic phenotype remains unclear. Here, we integrate structural, biophysical, and evolutionary analyses to systematically evaluate features associated with single amino acid substitutions across 24 human protein-based blood group systems. We analyse 319 variants with curated phenotypic annotations alongside 481 control variants, identifying key determinants of null and antigenic phenotypes. Null variants are characterized by high evolutionary conservation, burial within the protein core, loss of hydrophobicity, increased polarity, and a propensity for arginine substitutions. Antigenic variants are also enriched in arginine; however, in contrast to null variants, they tend to occur at less conserved, more solvent-accessible, and structurally flexible sites. Supervised machine learning models trained on structural and biophysical descriptors were applied to distinguish (i) null and (ii) antigenic variants from controls, achieving balanced accuracies of 0.82 and 0.63, respectively. Feature importance analysis identified predicted pathogenicity, solvent accessibility, and evolutionary conservation as the most predictive determinants of null variants, whereas hydrophobicity, conservation, and flexibility dominated antigen prediction. This work establishes a framework linking molecular variation to blood group phenotypes and provides a foundation for predicting the impact of novel missense mutations in transfusion medicine and beyond.

We investigate whether representations from AINN-P1--a protein foundation model trained autoregressively on tens of millions of natural protein sequences--transfer to the task of ranking antibody-antigen pairs by binding affinity. Casting affinity maturation as a learning-to-rank problem over the change in binding free energy ({Delta}{Delta}G), we compare a task-specific sequence model trained end-to-end from scratch against lightweight downstream heads built on top of frozen AINN-P1 embeddings, all evaluated under an identical five-fold cross-validation protocol. A regularized linear probe on the frozen embeddings already surpasses the from-scratch baseline, and an optimized lightweight head raises the mean Spearman rank correlation from 0.42 to 0.53--a relative improvement of approximately 28%-- while training in seconds and without any fine-tuning of the foundation model. Because a linear probe alone exceeds the fully trained end-to-end baseline, the gain is attributable to representation quality rather than to added downstream-model capacity. These results position frozen foundation-model embeddings as a strong, data-efficient default for affinity ranking in antibody engineering and establish a conservative lower bound that task-adaptive fine-tuning is expected to exceed.

As antimicrobial resistance outpaces antibiotic development, antimicrobial peptides (AMPs) have emerged as a promising class of alternative antibacterials, and computational predictors are increasingly used to prioritize AMP candidates. Such predictors are typically evaluated on binary AMP/non-AMP classification, which does not test whether they can identify peptides with clinically relevant potency against specific pathogens. We present BATTLE-AMP, a benchmarking framework that evaluates AMP predictors against experimentally measured minimum inhibitory concentrations (MICs) across clinically relevant bacterial species and strains. We surveyed 48 published methods, finding fewer than 25% reproducible, and benchmarked 10 model families (21 variants) using experimental MIC data, synthetic sequence perturbations, activity cliff analyses, and all-atom molecular dynamics (MD) simulations. Four findings emerge: (i) models trained on MIC data outperform binary classifiers regardless of architecture; (ii) the best model depends on the target pathogen, so model selection must be guided by the biological question; (iii) most models cannot distinguish active peptides from inactive sequences with identical amino acid composition; and (iv) activity cliffs remain unresolved by both machine learning and MD, marking a limit of current computational methods. BATTLE-AMP is released as an open Snakemake framework at https://github.com/szczurek-lab/battleamp-snakemake for benchmarking new models and scoring novel candidate libraries.

Glycans coat the surface of all cells, and every glycan is recognised by specific glycan-binding pro-teins (GBPs). There are no general tools that can accurately estimate the binding strength between glycan and GBP from the amino acid sequence of the GBP and the molecular structure of the glycan, represented as SMILES string. We describe models for predicting such binding strengths developed as a part of a Capstone Course at the University of Alberta. The models are trained on a dataset that combines BindingDB, a published database of small-molecule protein interactions, and data from glycan arrays measured by Consortium of Functional Glycomics (CFG). In this hybrid dataset of protein-ligand interactions the ligands are both glycans from CFG and small molecules from BindingDB; similarly, proteins include GBP and proteins from BindingDB. Three models are presented (i) ProMax which fuses ESM-2, MolFormer, and MolCLR features; (ii) APEX which constrains learning to a predetermined form, a physical model of binding; (iii) UltraMax adds inter-atomic distances for the ligands. To address the dataset's severe long-tail distribution, the models employ tail-aware losses for rare high-binding instances. Trained and evaluated on approximately one million protein--ligand pairs using hold-out splits for unseen molecules, the three models provide a unified framework for quantitative glycan-protein binding prediction. We observed that learning glycan-protein binding is harder than the similar task of learning small-molecule-protein interactions. Simple mirror-inversion tests led us to postulate that insufficient use of chiral features is an important source of difficulty in learning these interactions.

Single cell RNA-seq (scRNA) has provided unprecedented resolution into cellular and clonal heterogeneity. Computational approaches have enabled recovery of differentiation dynamics, yet current approaches do not evaluate discontinuous differentiation processes present in malignant leukemia. To address these gaps, we developed SupeRJump: a jump-drift-diffusion based supervised cell-fate model (https://github.com/namwob44/SupeRJump/). We deploy this approach in human bone marrow, murine aging hematopoiesis, and lentivirally barcoded mouse models of acute myeloid leukemia. Our framework introduces a semi-supervised pseudotime strategy to fit a jump-drift-diffusion model and batch correction for lineage fate predictions from absorbing Markov chains. We introduce metrics to quantify cell skewness toward particular lineages, transitions through intermediate progenitor states toward terminally differentiated states, and discontinuous transition dynamics. We use these metrics to identify cells preferentially biased for differentiation, their underlying transcriptional networks, and gene programs responsible for differentiation discontinuity.

T cells are a key part of the adaptive immune system. Using their surface-bound T cell antigen receptors (TCRs), these cells scan peptides and other antigens presented to them by major histocompatibility complex molecules (MHCs) on the surface of cells, searching for abnormalities. Although determining the map between TCRs and their target antigens is of vital importance for the design of safe and effective T cell-based vaccines and therapeutics, decoding these interactions is challenging. Experimental methods are not scalable, and sequence-based computational methods have issues generalising to new antigens. The IMMREP25 benchmark of methods for predicting T cell antigen specificity showed that AlphaFold-based methods promise improved generalisation to novel antigens. However, the ability of structure prediction models to predict T cell antigen specificity has not been robustly evaluated previously. In this work, we characterise AlphaFolds ability to predict T cell antigen specificity. We created a pipeline for high-throughput prediction of TCR:peptide-MHC (pMHC) structures using AlphaFold that is > 100 fold faster than the default implementation and used it to benchmark AlphaFold 3 (AF3) and similar models at predicting T cell antigen specificity. We investigated the underlying correlates of AlphaFold-derived binding scores and found that the models predictive power is related to the positioning of TCRs over the pMHC and not chemical interactions. Furthermore, we refine the AlphaFold-derived binding scores by training a machine learning model we call the PAE Aggregator. We then investigate AF3s ability to uncover the clustering rules of TCR repertoires and recapitulate mutational scanning experiments. These analyses show that AlphaFold3 clusters sequence-similar TCRs according to their binding mode and detects disrupting point mutations accurately. Our results highlight both the promise and the current limitations of structure-based approaches for predicting TCR specificity, guiding the development of more reliable immunological prediction methods.

AO_SCPLOWBSTRACTC_SCPLOWPeptide:MHC class I (pMHC-I) binding stability governs the persistence of antigenic complexes at the cell surface and plays a key role in facilitating downstream immunological signals such as antigen presentation, T-cell activation, and immunodominance. However, methods for in silico stability prediction remain underexplored relative to binding affinity prediction, in part because available half-life datasets are sparse and expensive to collect. Here, we perform a systematic reassessment of pMHC-I stability prediction using controlled, similarity-aware data splits and apply a recently introduced supervised transfer-learning strategy to MINT, an interaction-aware protein language model, pre-trained on binding affinity and fine-tuned for quantitative half-life prediction. We show that MINT improves stability prediction over standard ESM-2 representations and existing predictors, and that assay-conditioned recalibration corrects systematic shifts across experimental measurement modalities. Across eluted ligand, immunogenicity, and personalized neoantigen prioritization benchmarks, predicted stability provides signal beyond binding affinity, enriching for naturally presented and immunogenic peptides within affinity-filtered candidate sets. These results establish pMHC-I half-life as an orthogonal and transferable biophysical signal connecting peptide binding, surface presentation, and T-cell recognition, and provide a leakage-aware, assay-aware framework for future antigen-presentation modeling.

MotivationThe Functional Expansion of Specific T cell (FEST)-based assays combine short-term peptide stimulation with TCR sequencing to identify clonotypes that expand in response to specific antigens. These approaches have proven invaluable for detecting neoantigen-specific T cell responses, guiding vaccine development, and assessing checkpoint blockade efficacy. However, variability introduced by biological and technical replicates poses challenges for reproducibility and interpretation, and existing computational tools do not address replicate-level analysis in these assays. ResultsWe developed replicateFest, a computational framework implemented as an R package and Shiny web application, to analyze FEST-based TCR-seq data with and without replicates. replicateFest applies Fishers exact test for non-replicate datasets and negative binomial modeling for replicate experiments, returning adjusted p-values and odds ratios to identify clonotypes significantly expanded in antigen-stimulated conditions. The framework distinguishes FEST-expanded clonotypes (relative to a no-antigen control) and FEST-positive clonotypes (expanded compared to all other conditions). Validation using synthetic datasets confirmed accurate detection of antigen-specific clonotypes. Application to published HIV-1 epitope stimulation data reproduced original findings and demonstrated replicateFests utility for reproducibility assessment and quality control. Availability and ImplementationreplicateFest is freely available under the Apache-2.0 license as an R package at https://github.com/OncologyQS/replicateFest and as an interactive Shiny application at http://www.stat-apps.onc.jhmi.edu/FEST/.

Advancements in immunogenomics and immuno-oncology have enabled the development of personalized cancer vaccines (PCVs) that target cancer cell-specific somatic variants. A subset of these variants produce neoantigens that, when presented on tumor cells by MHC molecules, have the potential to elicit a robust and specific immune response. To date, there are over one hundred interventional studies listed on clinicaltrials.gov that explore the use of PCVs. We have supported a number of these trials through the creation of bioinformatic pipelines, tools, and procedures for the identification of patient-specific neoantigen candidates. While many of these steps have been automated, the final selection of neoantigen candidates often relies on expert manual review, creating a bottleneck that limits scalability and full automation of PCV workflows. Addressing this challenge, we introduce NEAT (Neoantigen Evaluation & Automated Triage), a machine learning-based approach that enables automated neoantigen candidate prioritization and supports the transition toward more scalable and reproducible PCV design. We implemented a prediction model trained and tested on existing vaccine design results from 33 patients and 1,943 peptides, across 3 clinical trials, including 439 peptides prioritized for PCV inclusion. This model uses features such as tumor variant allele frequency, RNA expression, driver gene status, binding/presentation scores, and transcript support level to automatically predict whether a peptide will be accepted, rejected, or require further human review before inclusion in a vaccine. The model achieved a sensitivity of 0.847 and specificity of 0.924, with an area under the curve of 0.955. The model predictions have been incorporated in pVACtools version 7. By integrating this model into the vaccine development pipeline, we foresee a significant reduction in the time required to transition from patient sample collection to vaccine manufacturing, thereby enhancing the efficiency and scalability of PCV production.