Using the antibody-antigen binding interface to train image-based deep neural networks for antibody-epitope classification.

High-throughput B-cell sequencing has opened up new avenues for investigating complex mechanisms underlying our adaptive immune response. These technological advances drive data generation and the need to mine and analyze the information contained in these large datasets, in particular the identification of therapeutic antibodies (Abs) or those associated with disease exposure and protection. Here, we describe our efforts to use artificial intelligence (AI)-based image-analyses for prospective classification of Abs based solely on sequence information. We hypothesized that Abs recognizing the same part of an antigen share a limited set of features at the binding interface, and that the binding site regions of these Abs share share common structure and physicochemical property patterns that can serve as a "fingerprint" to recognize uncharacterized Abs. We combined large-scale sequence-based protein-structure predictions to generate ensembles of 3-D Ab models, reduced the Ab binding interface to a 2-D image (fingerprint), used pre-trained convolutional neural networks to extract features, and trained deep neural networks (DNNs) to classify Abs. We evaluated this approach using Ab sequences derived from human HIV and Ebola viral infections to differentiate between two Abs, Abs belonging to specific B-cell family lineages, and Abs with different epitope preferences. In addition, we explored a different type of DNN method to detect one class of Abs from a larger pool of Abs. Testing on Ab sets that had been kept aside during model training, we achieved average prediction accuracies ranging from 71-96% depending on the complexity of the classification task. The high level of accuracies reached during these classification tests suggests that the DNN models were able to learn a series of structural patterns shared by Abs belonging to the same class. The developed methodology provides a means to apply AI-based image recognition techniques to analyze high-throughput B-cell sequencing datasets (repertoires) for Ab classification.

Identification of antigenic epitopes of thyroperoxidase, thyroglobulin and interleukin-24. Exploration of cross-reactivity with environmental allergens and possible role in urticaria and hypothyroidism.

BACKGROUND: Human proteins such as interleukin-24 (IL24), thyroperoxidase (TPO) and thyroglobulin (Tg) are targets of IgE or IgG autoantibodies. Why these proteins are recognized by autoantibodies in some patients with chronic spontaneous urticaria (CSU) or hypothyroidism is unknown. OBJECTIVE: Through in silico analysis, identify antigen patches of TPO, Tg and IL24 and compare the sequences of these human proteins with some prevalent allergens. METHODS: The amino acids sequences of IL24, thyroperoxidase and thyroglobulin were compared between them and with 22 environmental allergens. Phylogenetic studies and multiple pairing were carried out to explore the degree of protein identity and cover. The proteins without 3D structure reported in the database, were modeled by homology with "Swiss Modeller" and compared through PYMOL. Residues conserved and accessible to the solvent (rASA> 0.25) were located in the 3D model to identify possible areas of cross-reactivity and antigen binding. RESULTS: We build a 3D model of the TPO and thyroglobulin protein base on proteins closely related. Five epitopes for TPO, six for IL24 and six for thyroglobulin were predicted. The amino acid sequences of allergens from different sources (Dermatophagoides pteronyssinus, Blomia tropicalis, Betula verrucosa, Cynodon dactylon, Aspergillus fumigatus, Canis domesticus, Felis domesticus) were compared with human TPO, Tg and IL24. The cover and alignments between allergens and human proteins were low. CONCLUSION: We identify possible linear and conformational epitopes of TPO, Tg and IL24 that could be the target of IgE or IgG binding in patients with urticaria or hypothyroidism; These epitopes do not appear to be present among common environmental allergens, suggesting that autoreactivity to these human proteins are not by cross-reactivity.

Protein Barcodes Enable High-Dimensional Single-Cell CRISPR Screens.

CRISPR pools are being widely employed to identify gene functions. However, current technology, which utilizes DNA as barcodes, permits limited phenotyping and bulk-cell resolution. To enable novel screening capabilities, we developed a barcoding system operating at the protein level. We synthesized modules encoding triplet combinations of linear epitopes to generate >100 unique protein barcodes (Pro-Codes). Pro-Code-expressing vectors were introduced into cells and analyzed by CyTOF mass cytometry. Using just 14 antibodies, we detected 364 Pro-Code populations; establishing the largest set of protein-based reporters. By pairing each Pro-Code with a different CRISPR, we simultaneously analyzed multiple phenotypic markers, including phospho-signaling, on dozens of knockouts. Pro-Code/CRISPR screens found two interferon-stimulated genes, the immunoproteasome component Psmb8 and a chaperone Rtp4, are important for antigen-dependent immune editing of cancer cells and identified Socs1 as a negative regulator of Pd-l1. The Pro-Code technology enables simultaneous high-dimensional protein-level phenotyping of 100s of genes with single-cell resolution.

Biopanning of allergens from wasp sting patients.

OBJECTIVE: Wasp venom is a potentially important natural drug, but it can cause hypersensitivity reactions. The purpose of the present study was to systematically study the epitopes of wasp venom. METHODS: Using a random 12-peptide phage library, we performed antibody-binding epitope panning on ten serum samples from wasp sting victims at 3 h and 4 days after the sting. The panning epitopes were identified by high-throughput sequencing and matched with wasp venom proteins by BLAST. The panned antibody-binding epitopes were verified by ELISA. RESULTS: A total of 35 specific potential wasp venom epitopes in 4 days were identified. Amongst them, twelve peptide epitopes were matched with nine wasp venom proteins, namely, vitellogenin precursor, hexamerin 70b precursor, venom carboxylesterase-6 precursor, MRJP5, major royal jelly protein 8 precursor, venom acid phosphatase Acph-1 precursor, phospholipase A2, venom serine protease 34 precursor, and major royal jelly protein 9 precursor. The changes in serum IgM antibodies induced by wasp venom were confirmed by ELISA based on the 12 peptide epitopes. CONCLUSION: The nine wasp venom proteins are potential allergens, which should be excluded or modified in the potential biomedical applications of wasp venom.

Cross-modality deep learning-based prediction of TAP binding and naturally processed peptide.

Epitopes presented on MHC class I molecules pass multiple processing stages before their presentation on MHC molecules, the main ones being proteasomal cleavage and TAP binding. Transporter associated with antigen processing (TAP) binding is a necessary stage for most, but not all, MHC-I-binding peptides. The molecular determinants of TAP-binding peptides can be experimentally estimated from binding experiments and from the properties of peptides inducing a CD8 T cell response. We here propose novel optimization formalisms to combine binding and activation experimental results to produce a classifier for TAP binding using dual-output kernel and deep learning approaches. The application of these algorithms to the human and murine TAP binding leads to predictors that are much more precise than current state of the art methods. Moreover, the computed score is highly correlated with the observed binding energy. The new predictors show that TAP binding may be much more selective than previously assumed in humans and mice and sensitive to the properties of most positions of the peptides. Beyond the improved precision for TAP binding, we propose that the same approach holds in most molecular binding problems, where functional and binding measures are simultaneously available, and can be used to significantly improve the precision of binding prediction algorithms in general and immune system molecules specifically.

Characterization of novel Omp31 antigenic epitopes of Brucella melitensis by monoclonal antibodies.

BACKGROUND: Brucellosis is a severe zoonotic disease worldwide. Detection and identification of Brucella species are essential to prevent or treat brucellosis in humans and animals. The outer membrane protein-31 (Omp31) is a major protein of Brucellae except for B. abortus, while the Omp31 antigenic epitopes have not been extensively characterized yet. RESULTS: A total of 22 monoclonal antibodies (mAbs) were produced against Omp31 of Brucella (B.) melitensis, of which 13 recognized five linear epitopes, 7 reacted with semi-conformational epitopes and 2 reacted with conformational epitopes, respectively. The mAb isotypes were 11 (50%) IgG2a, 5 (23%) IgG1 and 6 (27%) IgM. On the basis of epitope recognition and reactivity levels, 8 mAbs including 3 IgM and 5 IgG clones were considered as highly reactive and potentially diagnostic antibodies. Among these mAbs, 7A3 (IgG1), 5B1 (IgG2a), 2C1 (IgG2a) and 5B3 (IgG2a) reacted with differently conserved linear epitopes of B. melitensis, B. ovis, B. suis and B. canis strains, while 5H3 (IgG2a) highly reacted with a conformational epitope of Omp31 when tested with several immunoassays. CONCLUSIONS: These potent monoclonal antibodies can be used for identifying Omp31 antigens or detecting B. melitensis and other Brucella species beyond B. abortus in vitro or in vivo.

Defining the immunogenicity and antigenicity of HLA epitopes is crucial for optimal epitope matching in clinical renal transplantation.

Transplantation of an human leukocyte antigen (HLA) mismatched graft can lead to the development of donor-specific antibodies (DSA), which can result in antibody mediated rejection and graft loss as well as complicate repeat transplantation. These DSA are induced by foreign epitopes present on the mismatched HLA antigens of the donor. However, not all epitopes appear to be equally effective in their ability to induce DSA. Understanding the characteristics of HLA epitopes is crucial for optimal epitope matching in clinical transplantation. In this review, the latest insights on HLA epitopes are described with a special focus on the definition of immunogenicity and antigenicity of HLA epitopes. Furthermore, the use of this knowledge to prevent HLA antibody formation and to select the optimal donor for sensitised transplant candidates will be discussed.

Monoclonal antibody binding-site diversity assessment with a cell-based clustering assay.

The diversity of a panel of antibodies that target a specific antigen can be established in various assay formats. In conventional epitope binning assays purified antibodies are tested in a pairwise manner: two antibodies that compete with each other for binding to an antigen are grouped into the same cluster or bin, while they are assigned to two different clusters when they do not compete. Here we present a high through put assay that enables grouping of crude hybridoma supernatants without a need for antibody purification. In addition, the assay does not require recombinant protein, because it is conducted on cells that express the antigen of interest. Hence, one can use the antibody-clustering assay for cell surface proteins that are not amenable to purification. Heavy chain variable region (VH) sequencing shows that VH composition within clusters is conserved. Finally, the assay is in good agreement with a conventional epitope binning assay with purified antigen.

Mapeo computacional de epítopos de células B presentes en el virus del dengue / B-cell epitopes mapping of dengue virus

Dengue es una infección viral de importancia en salud pública. El desarrollo de una vacuna va a depender del conocimiento de aquello epítopos que neutralicen la infección y por ello, se generaron e identificaron computacionalmente aquellos epítopos lineales consenso de células B presentes en el virus del dengue, a partir de los datos depositados en la base de datos Immune Epitope Database (IEDB). Posteriormente, se agruparon en bloques de ocho aminoácidos por cada serotipo mediante el programa Nomad, y cada uno de ellos fue reevaluado con el programa BepiPred para determinar su antigenicidad. Se lograron postular 172 de los 1.239 obtenidos en la IEDB.

Dengue virus has emerged as a major public health problem. An epitope-based approach to the development of vaccines, and for this reason, based on Immune Epitope Database database (IEDB), linear consensus B-cell epitopes were computationally generated and identified from dengue virus. Then, amino acids were grouped into blocks of eight amino acids per serotype through Nomad program. Each block was evaluated using BepiPred for determining antigenic prediction. 172 out of 1239 were obtained from IEDB.

Cost sensitive hierarchical document classification to triage PubMed abstracts for manual curation.

BACKGROUND: The Immune Epitope Database (IEDB) project manually curates information from published journal articles that describe immune epitopes derived from a wide variety of organisms and associated with different diseases. In the past, abstracts of scientific articles were retrieved by broad keyword queries of PubMed, and were classified as relevant (curatable) or irrelevant (not curatable) to the scope of the database by a Naïve Bayes classifier. The curatable abstracts were subsequently manually classified into categories corresponding to different disease domains. Over the past four years, we have examined how to further improve this approach in order to enhance classification performance and to reduce the need for manual intervention. RESULTS: Utilizing 89,884 abstracts classified by a domain expert as curatable or uncuratable, we found that a SVM classifier outperformed the previously used Naïve Bayes classifier for curatability predictions with an AUC of 0.899 and 0.854, respectively. Next, using a non-hierarchical and a hierarchical application of SVM classifiers trained on 22,833 curatable abstracts manually classified into three levels of disease specific categories we demonstrated that a hierarchical application of SVM classifiers outperformed non-hierarchical SVM classifiers for categorization. Finally, to optimize the hierarchical SVM classifiers' error profile for the curation process, cost sensitivity functions were developed to avoid serious misclassifications. We tested our design on a benchmark dataset of 1,388 references and achieved an overall category prediction accuracy of 94.4%, 93.9%, and 82.1% at the three levels of categorization, respectively. CONCLUSIONS: A hierarchical application of SVM algorithms with cost sensitive output weighting enabled high quality reference classification with few serious misclassifications. This enabled us to significantly reduce the manual component of abstract categorization. Our findings are relevant to other databases that are developing their own document classifier schema and the datasets we make available provide large scale real-life benchmark sets for method developers.

Bacteria modulate the CD8+ T cell epitope repertoire of host cytosol-exposed proteins to manipulate the host immune response.

The main adaptive immune response to bacteria is mediated by B cells and CD4+ T-cells. However, some bacterial proteins reach the cytosol of host cells and are exposed to the host CD8+ T-cells response. Both gram-negative and gram-positive bacteria can translocate proteins to the cytosol through type III and IV secretion and ESX-1 systems, respectively. The translocated proteins are often essential for the bacterium survival. Once injected, these proteins can be degraded and presented on MHC-I molecules to CD8+ T-cells. The CD8+ T-cells, in turn, can induce cell death and destroy the bacteria's habitat. In viruses, escape mutations arise to avoid this detection. The accumulation of escape mutations in bacteria has never been systematically studied. We show for the first time that such mutations are systematically present in most bacteria tested. We combine multiple bioinformatic algorithms to compute CD8+ T-cell epitope libraries of bacteria with secretion systems that translocate proteins to the host cytosol. In all bacteria tested, proteins not translocated to the cytosol show no escape mutations in their CD8+ T-cell epitopes. However, proteins translocated to the cytosol show clear escape mutations and have low epitope densities for most tested HLA alleles. The low epitope densities suggest that bacteria, like viruses, are evolutionarily selected to ensure their survival in the presence of CD8+ T-cells. In contrast with most other translocated proteins examined, Pseudomonas aeruginosa's ExoU, which ultimately induces host cell death, was found to have high epitope density. This finding suggests a novel mechanism for the manipulation of CD8+ T-cells by pathogens. The ExoU effector may have evolved to maintain high epitope density enabling it to efficiently induce CD8+ T-cell mediated cell death. These results were tested using multiple epitope prediction algorithms, and were found to be consistent for most proteins tested.

Chimpanzees use more varied receptors and ligands than humans for inhibitory killer cell Ig-like receptor recognition of the MHC-C1 and MHC-C2 epitopes.

Humans and chimpanzees have orthologous MHC class I, but few orthologous killer cell Ig-like receptors (KIR). Most divergent are lineage III KIR, which in humans include the inhibitory KIR2DL1 and 2DL2/3 specific for HLA-C. Six lineage III chimpanzee KIR were identified as candidate inhibitory MHC-C receptors and studied using cytolytic assays, to assess the capacity of a defined KIR to function with a defined MHC class I allotype, and direct binding assays with KIR-Fc fusion proteins. Pt-KIR2DL6 and 2DL8 were demonstrated to be inhibitory C1 receptors with a specificity and specificity-determining residue (lysine 44) like KIR2DL3. Analogously, Pt-KIR2DL7 is like KIR2DL1, an inhibitory C2 receptor having methionine 44. Pt-KIR3DL4 and 3DL5 are unusual lineage III KIR with D0 domains, which are also inhibitory C2 receptors with methionine 44. Removal of D0 from KIR3DL, or its addition to KIR2DL, had no effect on KIR function. Pt-KIR2DL9, a fourth inhibitory C2 receptor, has glutamate 44, a previously uncharacterized specificity-determining residue that is absent from human KIR. Reconstruction of the ancestral hominoid KIR sequence shows it encoded lysine 44, indicating that KIR having methionine 44 and glutamate 44 subsequently evolved by independent point substitutions. Thus, MHC-C2-specific KIR have evolved independently on at least two occasions. None of the six chimpanzee KIR studied resembles KIR2DL2, which interacts strongly with C1 and cross-reacts with C2. Whereas human HLA-B allotypes that have functional C1 epitopes are either rare (HLA-B*73) or geographically localized (HLA-B*46), some 25% of Patr-B allotypes have the C1 epitope and are functional KIR ligands.

Development of an epitope conservancy analysis tool to facilitate the design of epitope-based diagnostics and vaccines.

BACKGROUND: In an epitope-based vaccine setting, the use of conserved epitopes would be expected to provide broader protection across multiple strains, or even species, than epitopes derived from highly variable genome regions. Conversely, in a diagnostic and disease monitoring setting, epitopes that are specific to a given pathogen strain, for example, can be used to monitor responses to that particular infectious strain. In both cases, concrete information pertaining to the degree of conservancy of the epitope(s) considered is crucial. RESULTS: To assist in the selection of epitopes with the desired degree of conservation, we have developed a new tool to determine the variability of epitopes within a given set of protein sequences. The tool was implemented as a component of the Immune Epitope Database and Analysis Resources (IEDB), and is directly accessible at http://tools.immuneepitope.org/tools/conservancy. CONCLUSION: An epitope conservancy analysis tool was developed to analyze the variability or conservation of epitopes. The tool is user friendly, and is expected to aid in the design of epitope-based vaccines and diagnostics.

Automating document classification for the Immune Epitope Database.

BACKGROUND: The Immune Epitope Database contains information on immune epitopes curated manually from the scientific literature. Like similar projects in other knowledge domains, significant effort is spent on identifying which articles are relevant for this purpose. RESULTS: We here report our experience in automating this process using Naïve Bayes classifiers trained on 20,910 abstracts classified by domain experts. Improvements on the basic classifier performance were made by a) utilizing information stored in PubMed beyond the abstract itself b) applying standard feature selection criteria and c) extracting domain specific feature patterns that e.g. identify peptides sequences. We have implemented the classifier into the curation process determining if abstracts are clearly relevant, clearly irrelevant, or if no certain classification can be made, in which case the abstracts are manually classified. Testing this classification scheme on an independent dataset, we achieve 95% sensitivity and specificity in the 51.1% of abstracts that were automatically classified. CONCLUSION: By implementing text classification, we have sped up the reference selection process without sacrificing sensitivity or specificity of the human expert classification. This study provides both practical recommendations for users of text classification tools, as well as a large dataset which can serve as a benchmark for tool developers.

Classification of Proteus mirabilis TG 115 and CCUG 10701 into the Proteus O23 serogroup based on chemical and serological studies of O-polysaccharides.

INTRODUCTION: Bacteria of the genus Proteus are a common cause of urinary tract infections. The O-polysaccharide (OPS) chain of their lipopolysaccharide (LPS) defines the serological specificity of strains. Based on the OPS structures and the immunospecificity of the LPS, Proteus strains have been classified into 74 O-serogroups. MATERIALS AND METHODS: The OPS of P. mirabilis TG 115 was obtained by mild acid degradation of the LPS and studied by (1)H and (13)C nuclear magnetic resonance spectroscopy. Antisera were raised by immunization of rabbits with heat-killed bacteria. Serological studies were performed using enzyme immunosorbent assay, passive immunoheamolysis, inhibition experiments, absorption of O-antisera, and Western blot. RESULTS: The following structure of the P. mirabilis TG 115 OPS was established: --> 2)-beta-D-GalpA-(1--> 3)-alpha-D-GalpNAc-(1--> 4)-alpha-D-GalpA-(1--> 3)-beta-D-GlcpNAc-(1--> The same structure has been reported previously for the O-polysaccharides of P. mirabilis CCUG 10701 (O74) and P. mirabilis 41/57 (O23), except that they contain O-acetyl groups in non-stoichiometric quantities. Serological studies showed the antigenic identity of the three strains and their close serological relatedness to P. vulgaris 44/57. CONCLUSIONS: Based on the OPS structures and serological data, it is suggested to classify P. mirabilis 41/57, TG 115, and CCUG 10701 into one subgroup and P. mirabilis 42/57 and P. vulgaris 43/57 and 44/57 into another subgroup of the Proteus O23 serogroup.

IgE-reactive carbohydrate epitopes--classification, cross-reactivity, and clinical impact.

A glycan-related IgE-reactivity has been demonstrated in most allergen sources, especially in plant kingdom. Recent progress in glycobiology has allowed a clearer classification of these glyco-epitopes. Unlike classical peptide chain-based epitopes, glyco-epitopes can share significant structural homologies beyond the limits of protein families. These glycoepitopes are thus prone to extensive cross-reactivity. They have been called Cross-reactive Carbohydrate Determinants or CCD. Many of these glyco-epitopes behave as "panepitopes", leading to cross-reactivity between products as distant as pollens and hymenoptera venoms. But CCD are not universally cross-reactive and they rather cluster into subgroups such as plant CCD or fungal CCD. Because a monovalent IgE-binding is sufficient in serum-based assays, glyco-epitopes and CCD are classically considered as a potential source of positive in vitro results without clinical significance. But some authors recently demonstrated that glyco-epitopes could induce a response at the cell level and suggested that they might play a role in vivo. As long as in vitro assays include glycan- and CCD-related IgE responses, laboratory results should be carefully interpreted in the light of existing knowledge about the glycomes of natural products. IgE-reactivity of the patient's serum can also be tested towards a glycoprotein model such as bromelain.

The antigenic tip GPGRAFY of the V3 loop on HIV-1 gp120: genetic variability and subtypes.

V3 loop on HIV-1 gp120 is tightly correlated with syncytium formation, coreceptor usage, virus infectivity and antibody neutralization. The antigenic tip GPGRAFY with its flanking sequence has a conserved secondary structure, and is the target of neutralizing antibodies. We analyzed its genetic variability in 30096 M-group isolates and 269 O-group isolates. Subtype-related restricted mutations were observed, which could help to identify subtypes.

Monoclonal antibodies that distinguish antigenic variants of canine parvovirus.

Canine parvovirus (CPV) is classified as a member of the feline parvovirus (FPV) subgroup. CPV isolates are divided into three antigenic types: CPV type 2 (CPV-2), CPV-2a, and CPV-2b. Recently, new antigenic types of CPV were isolated from Vietnamese leopard cats and designated CPV-2c(a) or CPV-2c(b). CPV-2c viruses were distinguished from the other antigenic types of the FPV subgroup by the absence of reactivity with several monoclonal antibodies (MAbs). To characterize the antigenicity of CPV-2c, a panel of MAbs against CPV-2c was generated and epitopes recognized by these MAbs were examined by selection of escape mutants. Four MAbs were established and classified into three groups on the basis of their reactivities: MAbs which recognize CPV-2a, CPV-2b, and CPV-2c (MAbs 2G5 and 20G4); an MAb which reacts with only CPV-2b and CPV-2c(b) (MAb 21C3); and an MAb which recognizes all types of the FPV subgroup viruses (MAb 19D7). The reactivity of MAb 20G4 with CPV-2c was higher than its reactivities with CPV-2a and CPV-2b. These types of specificities of MAbs have not been reported previously. A mapping study by analysis of neutralization-resistant mutants showed that epitopes recognized by MAbs 21C3 and 19D7 belonged to antigenic site A. Substitution of the residues in site B and the other antigenic site influenced the epitope recognized by MAb 2G5. It was suggested that the epitope recognized by MAb 20G4 was related to antigenic site B. These MAbs are expected to be useful for the detection and classification of FPV subgroup isolates.