Search

Active learning allows algorithms to steer iterative experimentation to accelerate and de-risk molecular optimizations, but actively trained models might still exhibit poor performance during early project stages where the training data is limited and model exploitation might lead to analog identification with limited scaffold diversity. Here, we present ActiveDelta, an adaptive approach that leverages paired molecular representations to predict improvements from the current best training compound to prioritize further data acquisition. We apply the ActiveDelta concept to both graph-based deep (Chemprop) and tree-based (XGBoost) models during exploitative active learning for 99 Ki benchmarking datasets. We show that both ActiveDelta implementations excel at identifying more potent inhibitors compared to the standard exploitative active learning implementations of Chemprop, XGBoost, and Random Forest. The ActiveDelta approach is also able to identify more chemically diverse inhibitors in terms of their Murcko scaffolds. Finally, deep models such as Chemprop trained on data selected through ActiveDelta approaches can more accurately identify inhibitors in test data created through simulated time-splits. Overall, this study highlights the large potential for molecular pairing approaches to further improve popular active learning strategies in low data regimes by enabling faster and more accurate identification of more diverse molecular hits against critical drug targets.

Abstract The UV-mediated peptide exchange has enabled the generation of multiple different MHC multimer specificities in parallel, surpassing tedious individual refolding of MHC molecules with peptide ligands. Murine models are acknowledged as an effective tool for preclinical research to advance our understanding of immunological mechanisms, with the potential translatability of key learnings from mouse models to the clinic. The common inbred mouse strain BALB/c is frequently used in immunological research. However, for the BALB/c histocompatibility (H)-2 alleles availability of conditional ligand has been limited. To overcome this challenge, we design and experimentally validate conditional ligands restricted to murine MHC class I alleles H2Dd and H2Kd. In addition, we demonstrate the ability of the three H2d molecules and two additional C57BL/6 H2b molecules folded in-house with conditional ligands to generate fluorescently labeled peptide-H2 tetramers that allow staining of antigen-specific CD8+ T cells in splenocyte samples. Finally, we generate large peptide-H-2 multimer libraries with a DNA-barcode labeling system for high-throughput interrogation of CD8+ T cell specificity in murine splenocyte samples. Consequently, the described techniques will contribute to our understanding of the antigen-specific CD8+ T cell repertoire in murine preclinical models of various diseases.

Abstract Background We have recently shown extensive sequence and conformational homology between tumor-associated antigens (TAAs) and antigens derived from microorganisms (MoAs). The present study aimed to assess the breadth of T-cell recognition specific to MoAs and the corresponding TAAs in healthy subjects (HS) and patients with cancer (CP). Method A library of > 100 peptide-MHC (pMHC) combinations was used to generate DNA-barcode labelled multimers. Homologous peptides were selected from the Cancer Antigenic Peptide Database, as well as Bacteroidetes/Firmicutes-derived peptides. They were incubated with CD8 + T cells from the peripheral blood of HLA-A*02:01 healthy individuals (n = 10) and cancer patients (n = 16). T cell recognition was identified using tetramer-staining analysis. Cytotoxicity assay was performed using as target cells TAP-deficient T2 cells loaded with MoA or the paired TuA. Results A total of 66 unique pMHC recognized by CD8+ T cells across all groups were identified. Of these, 21 epitopes from microbiota were identified as novel immunological targets. Reactivity against selected TAAs was observed for both HS and CP. pMHC tetramer staining confirmed CD8+ T cell populations cross-reacting with CTA SSX2 and paired microbiota epitopes. Moreover, PBMCs activated with the MoA where shown to release IFNγ as well as to exert cytotoxic activity against cells presenting the paired TuA. Conclusions Several predicted microbiota-derived MoAs are recognized by T cells in HS and CP. Reactivity against TAAs was observed also in HS, primed by the homologous bacterial antigens. CD8+ T cells cross-reacting with MAGE-A1 and paired microbiota epitopes were identified in three subjects. Therefore, the microbiota can elicit an extensive repertoire of natural memory T cells to TAAs, possibly able to control tumor growth (“natural anti-cancer vaccination”). In addition, non-self MoAs can be included in preventive/therapeutic off-the-shelf cancer vaccines with more potent anti-tumor efficacy than those based on TAAs.

Multi-factor screenings are commonly used in diverse applications in medicine and bioengineering, including optimizing combination drug treatments and microbiome engineering. Despite the advances in high-throughput technologies, large-scale experiments typically remain prohibitively expensive. Here we introduce a machine learning platform, structure-augmented regression (SAR), that exploits the intrinsic structure of each biological system to learn a high-accuracy model with minimal data requirement. Under different environmental perturbations, each biological system exhibits a unique, structured phenotypic response. This structure can be learned based on limited data and once learned, can constrain subsequent quantitative predictions. We demonstrate that SAR requires significantly fewer data comparing to other existing machine-learning methods to achieve a high prediction accuracy, first on simulated data, then on experimental data of various systems and input dimensions. We then show how a learned structure can guide effective design of new experiments. Our approach has implications for predictive control of biological systems and an integration of machine learning prediction and experimental design.

Active machine learning is an established and increasingly popular experimental design technique where the machine learning model can request additional data to improve the model's predictive performance. It is generally assumed that this data is optimal for the machine learning model since it relies on the model's predictions or model architecture and therefore cannot be transferred to other models. Inspired by research in pedagogy, we here introduce the concept of yoked machine learning where a second machine learning model learns from the data selected by another model. We found that in 48% of the benchmarked combinations, yoked learning performed similar or better than active learning. We analyze distinct cases in which yoked learning can improve active learning performance. In particular, we prototype yoked deep learning (YoDeL) where a classic machine learning model provides data to a deep neural network, thereby mitigating challenges of active deep learning such as slow refitting time per learning iteration and poor performance on small datasets. In summary, we expect the new concept of yoked (deep) learning to provide a competitive option to boost the performance of active learning and benefit from distinct capabilities of multiple machine learning models during data acquisition, training, and deployment.

Background Neoantigens can serve as targets for T cell-mediated antitumor immunity via personalized neopeptide vaccines. Interim data from our clinical study NCT03715985 showed that the personalized peptide-based neoantigen vaccine EVX-01, formulated in the liposomal adjuvant, CAF09b, was safe and able to elicit EVX-01-specific T cell responses in patients with metastatic melanoma. Here, we present results from the dose-escalation part of the study, evaluating the feasibility, safety, efficacy, and immunogenicity of EVX-01 in addition to anti-PD-1 therapy.Methods Patients with metastatic melanoma on anti-PD-1 therapy were treated in three cohorts with increasing vaccine dosages (twofold and fourfold). Tumor-derived neoantigens were selected by the AI platform PIONEER and used in personalized therapeutic cancer peptide vaccines EVX-01. Vaccines were administered at 2-week intervals for a total of three intraperitoneal and three intramuscular injections. The study’s primary endpoint was safety and tolerability. Additional endpoints were immunological responses, survival, and objective response rates.Results Compared with the base dose level previously reported, no new vaccine-related serious adverse events were observed during dose escalation of EVX-01 in combination with an anti-PD-1 agent given according to local guidelines. Two patients at the third dose level (fourfold dose) developed grade 3 toxicity, most likely related to pembrolizumab. Overall, 8 out of the 12 patients had objective clinical responses (6 partial response (PR) and 2 CR), with all 4 patients at the highest dose level having a CR (1 CR, 3 PR). EVX-01 induced peptide-specific CD4+ and/or CD8+T cell responses in all treated patients, with CD4+T cells as the dominating responses. The magnitude of immune responses measured by IFN-γ ELISpot assay correlated with individual peptide doses. A significant correlation between the PIONEER quality score and induced T cell immunogenicity was detected, while better CRs correlated with both the number of immunogenic EVX-01 peptides and the PIONEER quality score.Conclusion Immunization with EVX-01-CAF09b in addition to anti-PD-1 therapy was shown to be safe and well tolerated and elicit vaccine neoantigen-specific CD4+and CD8+ T cell responses at all dose levels. In addition, objective tumor responses were observed in 67% of patients. The results encourage further assessment of the antitumor efficacy of EVX-01 in combination with anti-PD-1 therapy.

BackgroundMutation-derived neoantigens are critical targets for tumor rejection in cancer immunotherapy, and better tools for neoepitope identification and prediction are needed to improve neoepitope targeting strategies. Computational tools have enabled the identification of patient-specific neoantigen candidates from sequencing data, but limited data availability has hindered their capacity to predict which of the many neoepitopes will most likely give rise to T cell recognition. MethodTo address this, we make use of experimentally validated T cell recognition towards 17,500 neoepitope candidates, with 467 being T cell recognized, across 70 cancer patients undergoing immunotherapy. ResultsWe evaluated 27 neoepitope characteristics, and created a random forest model, IMPROVE, to predict neoepitope immunogenicity. The presence of hydrophobic and aromatic residues in the peptide binding core were the most important features for predicting neoepitope immunogenicity.ConclusionOverall, IMPROVE was found to significantly advance the identification of neoepitopes compared to other current methods.

Abstract Established molecular machine learning models process individual molecules as inputs to predict their biological, chemical, or physical properties. However, such algorithms require large datasets and have not been optimized to predict property differences between molecules, limiting their ability to learn from smaller datasets and to directly compare the anticipated properties of two molecules. Many drug and material development tasks would benefit from an algorithm that can directly compare two molecules to guide molecular optimization and prioritization, especially for tasks with limited available data. Here, we develop DeepDelta, a pairwise deep learning approach that processes two molecules simultaneously and learns to predict property differences between two molecules from small datasets. On 10 ADMET benchmark tasks, our DeepDelta approach significantly outperforms two established molecular machine learning algorithms, the directed message passing neural network (D-MPNN) ChemProp and Random Forest using radial fingerprints, for 70% of benchmarks in terms of Pearson’s r, 60% of benchmarks in terms of mean absolute error (MAE), and all external test sets for both Pearson’s r and MAE. We further analyze our performance and find that DeepDelta is particularly outperforming established approaches at predicting large differences in molecular properties and can perform scaffold hopping. Furthermore, we derive mathematically fundamental computational tests of our models based on mathematical invariants and show that compliance to these tests correlates with overall model performance — providing an innovative, unsupervised, and easily computable measure of expected model performance and applicability. Taken together, DeepDelta provides an accurate approach to predict molecular property differences by directly training on molecular pairs and their property differences to further support fidelity and transparency in molecular optimization for drug development and the chemical sciences.

Abstract Background Kidney transplant recipients receive maintenance immunosuppressive therapy to avoid allograft rejection resulting in increased risk of infections and infection-related morbidity and mortality. Approximately 98% of adults are infected with varicella zoster virus, which upon reactivation causes herpes zoster. The incidence of herpes zoster is higher in kidney transplant recipients than in immunocompetent individuals, and kidney transplant recipients are at increased risk of severe herpes zoster-associated disease. Vaccination with adjuvanted recombinant glycoprotein E subunit herpes zoster vaccine (RZV) prevents herpes zoster in older adults with excellent efficacy (90%), and vaccination of kidney transplant candidates is recommended in Danish and international guidelines. However, the robustness and duration of immune responses after RZV vaccination, as well as the optimal timing of vaccination in relation to transplantation remain unanswered questions. Thus, the aim of this study is to characterize the immune response to RZV vaccination in kidney transplant candidates and recipients at different timepoints before and after transplantation. Methods The Herpes Virus Infections in Kidney Transplant Patients (HINT) study is a prospective observational cohort study. The study will include kidney transplant candidates on the waiting list for transplantation (n = 375) and kidney transplant recipients transplanted since January 1, 2019 (n = 500) from all Danish kidney transplant centers who are offered a RZV vaccine as routine care. Participants are followed with repeated blood sampling until 12 months after inclusion. In the case of transplantation or herpes zoster disease, additional blood samples will be collected until 12 months after transplantation. The immune response will be characterized by immunophenotyping and functional characterization of varicella zoster virus-specific T cells, by detection of anti-glycoprotein E antibodies, and by measuring cytokine profiles. Discussion The study will provide new knowledge on the immune response to RZV vaccination in kidney transplant candidates and recipients and the robustness and duration of the response, potentially enhancing preventive strategies against herpes zoster in a population at increased risk. Trial registration ClinicalTrials.gov (NCT05604911).

Abstract Background Patients undergoing cardiac surgery are prone to numerous complications. Increased vascular permeability may be associated with morbidity and mortality due to hemodynamic instability, fluid overload, and edema formation. We hypothesized that markers of endothelial injury and inflammation are associated with capillary leak, ultimately increasing the risk of postoperative complications. Methods In this prospective, observational, multidisciplinary cohort study at our tertiary academic medical center, we recruited 405 cardiac surgery patients. Patients were assessed daily using body impedance electrical analysis, ultrasound, sublingual intravital microscopy, and analysis of serum biomarkers. Multivariable models, as well as machine learning, were used to study the association of angiopoietin-2 with extracellular water as well as common complications after cardiac surgery. Results The majority of patients underwent coronary artery bypass grafting, valvular, or aortic surgeries. Across the groups, extracellular water increased postoperatively (20 ± 6 preoperatively to 29 ± 7L on postoperative day 2; P

Abstract In this study we performed a step-wise optimization of biologically active IL-2 for delivery using E. coli Nissle 1917. Engineering of the strain was coupled with an in vitro cell assay to measure the biological activity of microbially produced IL-2 (mi-IL2). Next, we assessed the immune modulatory potential of mi-IL2 using a 3D tumor spheroid model demonstrating a strong effect on immune cell activation. Finally, we evaluated the anticancer properties of the engineered strain in a murine CT26 tumor model. The engineered strain was injected intravenously and selectively colonized tumors. The treatment was well-tolerated, and tumors of treated mice showed a modest reduction in tumor growth rate, as well as significantly elevated levels of IL-2 in the tumor. This work demonstrates a workflow for researchers interested in engineering E. coli Nissle for a new class of microbial therapy against cancer.

Abstract Recent findings have positioned tumor mutation-derived neoepitopes as attractive targets for cancer immunotherapy. Cancer vaccines that deliver neoepitopes via various vaccine formulations have demonstrated promising preliminary results in patients and animal models. In the presented work, we assessed the ability of plasmid DNA to confer neoepitope immunogenicity and anti-tumor effect in two murine syngeneic cancer models. We demonstrated that neoepitope DNA vaccination led to anti-tumor immunity in the CT26 and B16F10 tumor models, with the long-lasting presence of neoepitope-specific T-cell responses in blood, spleen, and tumors after immunization. We further observed that engagement of both the CD4+ and CD8+ T cell compartments was essential to hamper tumor growth. Additionally, combination therapy with immune checkpoint inhibition provided an additive effect, superior to either monotherapy. DNA vaccination offers a versatile platform that allows the encoding of multiple neoepitopes in a single formulation and is thus a feasible strategy for personalized immunotherapy via neoepitope vaccination.

Abstract Background Life-long immunosuppressive treatment after liver transplantation (LT) prevents graft rejection but predisposes the LT recipient to infections. Herpesvirus infections are associated with morbidity and mortality among LT recipients. Among those, especially cytomegalovirus (CMV) and varicella-zoster virus (VZV) pose challenges after LT. The aim of this study is to provide an in-depth characterization of the cellular immune response against CMV and VZV infections in LT recipients and identify potential risk factors for infection. Methods The Herpesvirus Infections in Solid Organ Transplant Recipients – Liver Transplant Study (HISTORY) consists of an epidemiological and immunological substudy. The epidemiological substudy is a retrospective observational cohort study that includes all patients who underwent LT in Denmark between 2010 and 2023 (N ≈ 500). Using data from nationwide hospital records and national health registries, the incidence of and clinical risk factors for CMV and VZV infections will be determined. The immunological substudy is an explorative prospective observational cohort study including patients enlisted for LT in Denmark during a 1.5-year period (N > 80). Participants will be followed with scheduled blood samples until 12 months after LT. CMV- and VZV-derived peptides will be predicted for their likelihood to be presented in participants based on their HLA type. Peptide-MHC complexes (pMHC) will be produced to isolate CMV- and VZV-specific T cells from peripheral blood mononuclear cells before and after CMV and VZV infection. Their frequency, T cell receptor sequences, and phenotypic characteristics will be examined, and in a subset of participants, CMV- and VZV-specific T cells will be expanded ex vivo. Discussion This study will provide novel insight into T cell immunity required for viral control of CMV and VZV and has the potential to develop a prediction model to identify LT recipients at high risk for infection based on a combination of clinical and immunological data. Furthermore, this study has the potential to provide proof-of-concept for adoptive T cell therapy against CMV and VZV. Combined, this study has the potential to reduce the burden and consequence of CMV and VZV infections and improve health and survival in LT recipients. Trial registration ClinicalTrials.gov (NCT05532540), registered 8 September 2022.

Abstract Background Patients’ self-management of the implications of their disease(s) is becoming increasingly important. Research shows that hospitalization disrupts established self-management routines. Nurses can play an important role in supporting patients’ self-management. The aim of this study is to describe how nurses support the self-management of hospitalized patients through verbal communication during routine nursing care. Methods A qualitative descriptive study, using overt, non-participant observations was conducted on three wards of a general teaching hospital in the Netherlands. A total of 215 hours of nursing work during 49 shifts was observed. Data was analyzed using thematic analysis based on the six phases of Braun and Clarke. Results Our observations showed that nurses discuss patients’ self-management mainly in short conversations during the care provision. Nurses ask patients about their self-management at home and stimulate patients to express their opinions and to be involved in the care process. Three themes reflect how nurses support self-management: ‘Discussing patient’s self-management’, ‘Enhancing patient’s involvement in care’ and ‘Focusing on patient’s perspective’. Conclusion Hospital nurses have methods to support hospitalized patients’ self-management but it does not seem to be an integral part of daily practice. Given current developments in healthcare, it is reasonable to argue that self-management should be given greater emphasis within the hospital setting, requiring a collaborative approach with patients and other healthcare professionals across the care continuum.

Background Adoptive cell therapy (ACT) has shown promising results for the treatment of cancer and viral infections. Successful ACT relies on ex vivo expansion of large numbers of desired T-cells with strong cytotoxic capacity and in vivo persistence, which constitutes the greatest challenge to current ACT strategies. Here, in this study, we present a novel technology for ex vivo expansion of antigen-specific T-cells; artificial antigen-presenting scaffolds (Ag-scaffolds) consisting of a dextran-polysaccharide backbone, decorated with combinations of peptide-Major Histocompatibility Complex (pMHC), cytokines and co-stimulatory molecules, enabling coordinated stimulation of antigen-specific T-cells.Methods The capacity of Ag-scaffolds to expand antigen-specific T-cells was explored in ex vivo cultures with peripheral blood mononuclear cells from healthy donors and patients with metastatic melanoma. The resulting T-cell products were assessed for phenotypic and functional characteristics.Results We identified an optimal Ag-scaffold for expansion of T-cells for ACT, carrying pMHC and interleukin-2 (IL-2) and IL-21, with which we efficiently expanded both virus-specific and tumor-specific CD8+ T cells from peripheral blood of healthy donors and patients, respectively. The resulting T-cell products were characterized by a high frequency of antigen-specific cells with high self-renewal capacity, low exhaustion, a multifunctional cytokine profile upon antigen-challenge and superior tumor killing capacity. This demonstrates that the coordinated stimuli provided by an optimized stoichiometry of TCR engaging (pMHC) and stimulatory (cytokine) moieties is essential to obtain desired T-cell characteristics. To generate an ‘off-the-shelf’ multitargeting Ag-scaffold product of relevance to patients with metastatic melanoma, we identified the 30 most frequently recognized shared HLA-A0201-restricted melanoma epitopes in a cohort of 87 patients. By combining these in an Ag-scaffold product, we were able to expand tumor-specific T-cells from 60–70% of patients with melanoma, yielding a multitargeted T-cell product with up to 25% specific and phenotypically and functionally improved T cells.Conclusions Taken together, the Ag-scaffold represents a promising new technology for selective expansion of antigen-specific CD8+ T cells directly from blood, yielding a highly specific and functionally enhanced T-cell product for ACT.

Checkpoint inhibition (CPI) therapy and adoptive cell therapy with autologous tumor-infiltrating lymphocytes (TIL-based ACT) are the two most effective immunotherapies for the treatment of metastatic melanoma. While CPI has been the dominating therapy in the past decade, TIL-based ACT is beneficial for individuals even after progression on previous immunotherapies. Given that notable differences in response have been made when used as a subsequent treatment, we investigated how the qualities of TILs changed when the ex vivo microenvironment of intact tumor fragments were modulated with checkpoint inhibitors targeting programmed death receptor 1 (PD-1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4). Initially, we show that unmodified TILs from CPI-resistant individuals can be produced, are overwhelmingly terminally differentiated, and are capable of responding to tumor. We then investigate these properties in ex vivo checkpoint modulated TILs finding that that they retain these qualities. Lastly, we confirmed the specificity of the TILs to the highest responding tumor antigens, and identified this reactivity resides largely in CD39+CD69+ terminally differentiated populations. Overall, we found that anti-PD-1 will alter the proliferative capacity while anti-CTLA4 will influence breadth of antigen specificity.

Novel single-cell-based technologies hold the promise of matching T cell receptor (TCR) sequences with their cognate peptide-MHC recognition motif in a high-throughput manner. Parallel capture of TCR transcripts and peptide-MHC is enabled through the use of reagents labeled with DNA barcodes. However, analysis and annotation of such single-cell sequencing (SCseq) data are challenged by dropout, random noise, and other technical artifacts that must be carefully handled in the downstream processing steps. We here propose a rational, data-driven method termed ITRAP (improved T cell Receptor Antigen Paring) to deal with these challenges, filtering away likely artifacts, and enable the generation of large sets of TCR-pMHC sequence data with a high degree of specificity and sensitivity, thus outputting the most likely pMHC target per T cell. We have validated this approach across 10 different virus-specific T cell responses in 16 healthy donors. Across these samples, we have identified up to 1494 high-confident TCR-pMHC pairs derived from 4135 single cells.