DiscoTope-3.0: improved B-cell epitope prediction using inverse folding latent representations.

Accurate computational identification of B-cell epitopes is crucial for the development of vaccines, therapies, and diagnostic tools. However, current structure-based prediction methods face limitations due to the dependency on experimentally solved structures. Here, we introduce DiscoTope-3.0, a markedly improved B-cell epitope prediction tool that innovatively employs inverse folding structure representations and a positive-unlabelled learning strategy, and is adapted for both solved and predicted structures. Our tool demonstrates a considerable improvement in performance over existing methods, accurately predicting linear and conformational epitopes across multiple independent datasets. Most notably, DiscoTope-3.0 maintains high predictive performance across solved, relaxed and predicted structures, alleviating the need for experimental structures and extending the general applicability of accurate B-cell epitope prediction by 3 orders of magnitude. DiscoTope-3.0 is made widely accessible on two web servers, processing over 100 structures per submission, and as a downloadable package. In addition, the servers interface with RCSB and AlphaFoldDB, facilitating large-scale prediction across over 200 million cataloged proteins. DiscoTope-3.0 is available at: https://services.healthtech.dtu.dk/service.php?DiscoTope-3.0.

Comparative Structure Analysis of the Multi-Domain, Cell Envelope Proteases of Lactic Acid Bacteria.

Lactic acid bacteria (LAB) have an extracellular proteolytic system that includes a multi-domain, cell envelope protease (CEP) with a subtilisin homologous protease domain. These CEPs have different proteolytic activities despite having similar protein sequences. Structural characterization has previously been limited to CEP homologs of dairy- and human-derived LAB strains, excluding CEPs of plant-derived LAB strains. CEP structures are a challenge to determine experimentally due to their large size and attachment to the cell envelope. This study aims to clarify the prevalence and structural diversity of CEPs by using the structure prediction software AlphaFold 2. Domain boundaries are clarified based on a comparative analysis of 21 three-dimensional structures, revealing novel domain architectures of CEP homologs that are not necessarily restricted to specific LAB species or ecological niches. The C-terminal flanking region of the protease domain is divided into fibronectin type-III-like domains with various structural traits. The analysis also emphasizes the existence of two distinct domains for cell envelope attachment that are preceded by an intrinsically disordered cell wall spanning domain. The domain variants and their combinations provide CEPs with different stability, proteolytic activity, and potentially adhesive properties, making CEPs targets for steering proteolytic activity with relevance for both food development and human health.

Antioxidant peptides from alternative sources reduce lipid oxidation in 5% fish oil-in-water emulsions (pH 4) and fish oil-enriched mayonnaise.

Bioinformatics tools were used to predict radical scavenging and metal chelating activities of peptides derived from abundant potato, seaweed, microbial, and spinach proteins. The antioxidant activity was evaluated in 5% oil-in-water emulsions (pH4) and best-performing peptides were tested in mayonnaise and compared with EDTA. Emulsion physical stability was intact. The peptide DDDNLVLPEVYDQD showed the highest protection against oxidation in both emulsions by retarding the formation of oxidation products and depletion of tocopherols during storage, but it was less efficient than EDTA when evaluated in mayonnaise. In low-fat emulsions, formation of hydroperoxides was reduced 4-folds after 5 days compared to control. The concentration effect of the peptide was confirmed in mayonnaise at the EDTA equimolar concentration. The second-best performing peptides were NNKWVPCLEFETEHGFVYREHH in emulsion and AGDWLIGDR in mayonnaise. In general, the peptide efficacy was higher in low-fat emulsions. Results demonstrated that peptide negative net charge was important for chelating activity.

Signatures of T cell immunity revealed using sequence similarity with TCRDivER algorithm.

Changes in the T cell receptor (TCR) repertoires have become important markers for monitoring disease or therapy progression. With the rise of immunotherapy usage in cancer, infectious and autoimmune disease, accurate assessment and comparison of the "state" of the TCR repertoire has become paramount. One important driver of change within the repertoire is T cell proliferation following immunisation. A way of monitoring this is by investigating large clones of individual T cells believed to bind epitopes connected to the disease. However, as a single target can be bound by many different TCRs, monitoring individual clones cannot fully account for T cell cross-reactivity. Moreover, T cells responding to the same target often exhibit higher sequence similarity, which highlights the importance of accounting for TCR similarity within the repertoire. This complexity of binding relationships between a TCR and its target convolutes comparison of immune responses between individuals or comparisons of TCR repertoires at different timepoints. Here we propose TCRDivER algorithm (T cell Receptor Diversity Estimates for Repertoires), a global method of T cell repertoire comparison using diversity profiles sensitive to both clone size and sequence similarity. This approach allowed for distinction between spleen TCR repertoires of immunised and non-immunised mice, showing the need for including both facets of repertoire changes simultaneously. The analysis revealed biologically interpretable relationships between sequence similarity and clonality. These aid in understanding differences and separation of repertoires stemming from different biological context. With the rise of availability of sequencing data we expect our tool to find broad usage in clinical and research applications.

Widespread amyloidogenicity potential of multiple myeloma patient-derived immunoglobulin light chains.

BACKGROUND: In a range of human disorders such as multiple myeloma (MM), immunoglobulin light chains (IgLCs) can be produced at very high concentrations. This can lead to pathological aggregation and deposition of IgLCs in different tissues, which in turn leads to severe and potentially fatal organ damage. However, IgLCs can also be highly soluble and non-toxic. It is generally thought that the cause for this differential solubility behaviour is solely found within the IgLC amino acid sequences, and a variety of individual sequence-related biophysical properties (e.g. thermal stability, dimerisation) have been proposed in different studies as major determinants of the aggregation in vivo. Here, we investigate biophysical properties underlying IgLC amyloidogenicity. RESULTS: We introduce a novel and systematic workflow, Thermodynamic and Aggregation Fingerprinting (ThAgg-Fip), for detailed biophysical characterisation, and apply it to nine different MM patient-derived IgLCs. Our set of pathogenic IgLCs spans the entire range of values in those parameters previously proposed to define in vivo amyloidogenicity; however, none actually forms amyloid in patients. Even more surprisingly, we were able to show that all our IgLCs are able to form amyloid fibrils readily in vitro under the influence of proteolytic cleavage by co-purified cathepsins. CONCLUSIONS: We show that (I) in vivo aggregation behaviour is unlikely to be mechanistically linked to any single biophysical or biochemical parameter and (II) amyloidogenic potential is widespread in IgLC sequences and is not confined to those sequences that form amyloid fibrils in patients. Our findings suggest that protein sequence, environmental conditions and presence and action of proteases all determine the ability of light chains to form amyloid fibrils in patients.

Computational Modeling of Antibody and T-Cell Receptor (CDR3 Loops).

Antibodies and T-cell receptors have been a subject of much interest due to their central role in the immune system and their potential applications in several biotechnological and medical applications from cancer therapy to vaccine development. A unique feature of these two lymphocyte receptors is their ability to bind a huge variety of different (pathogen) targets. This ability stems from six short loops in the binding domain that have hypervariable sequence due to genetic recombination mechanism. Particularly one of these loops, the third complementarity determining region (CDR3), has the highest sequence variability and a dominant role in binding the target. However, it has also been proven the most difficult to be modeled structurally, which is vitally important for downstream tasks such as binding prediction. This difficulty stems from its variability in sequence that both reduces the possibility of finding homologues and introduces unique structural features in the loop. We present here a general protocol for modeling such loops in antibodies and T-cell receptors. We also discuss the difficulties in loop modeling and the advantages and limitations of different modeling methods.

A Computational Pipeline for Predicting Cancer Neoepitopes.

Mutant Peptide eXtractor and Informer (MuPeXI), by Bjerregaard et al. (Cancer Immunol Immunother CII 66:1123-1130, 2017), is a program which identifies tumor-specific peptides and assesses their potential to be neoepitopes. MuPeXI takes as input a VCF file and a list of human leukocyte antigen (HLA) types and optionally a gene expression profile to assess a peptide's potential to be a neoepitope. MuPeXI can be downloaded and run both locally and on a web server. Here, we describe a pipeline for processing the input data so that it can be used for MuPeXI and how to run MuPeXI both locally and as a web server.

BepiPred-3.0: Improved B-cell epitope prediction using protein language models.

B-cell epitope prediction tools are of great medical and commercial interest due to their practical applications in vaccine development and disease diagnostics. The introduction of protein language models (LMs), trained on unprecedented large datasets of protein sequences and structures, tap into a powerful numeric representation that can be exploited to accurately predict local and global protein structural features from amino acid sequences only. In this paper, we present BepiPred-3.0, a sequence-based epitope prediction tool that, by exploiting LM embeddings, greatly improves the prediction accuracy for both linear and conformational epitope prediction on several independent test sets. Furthermore, by carefully selecting additional input variables and epitope residue annotation strategy, performance was further improved, thus achieving unprecedented predictive power. Our tool can predict epitopes across hundreds of sequences in minutes. It is freely available as a web server and a standalone package at https://services.healthtech.dtu.dk/service.php?BepiPred-3.0 with a user-friendly interface to navigate the results.

NetSurfP-3.0: accurate and fast prediction of protein structural features by protein language models and deep learning.

Recent advances in machine learning and natural language processing have made it possible to profoundly advance our ability to accurately predict protein structures and their functions. While such improvements are significantly impacting the fields of biology and biotechnology at large, such methods have the downside of high demands in terms of computing power and runtime, hampering their applicability to large datasets. Here, we present NetSurfP-3.0, a tool for predicting solvent accessibility, secondary structure, structural disorder and backbone dihedral angles for each residue of an amino acid sequence. This NetSurfP update exploits recent advances in pre-trained protein language models to drastically improve the runtime of its predecessor by two orders of magnitude, while displaying similar prediction performance. We assessed the accuracy of NetSurfP-3.0 on several independent test datasets and found it to consistently produce state-of-the-art predictions for each of its output features, with a runtime that is up to to 600 times faster than the most commonly available methods performing the same tasks. The tool is freely available as a web server with a user-friendly interface to navigate the results, as well as a standalone downloadable package.

Antioxidant peptides derived from potato, seaweed, microbial and spinach proteins: Oxidative stability of 5% fish oil-in-water emulsions.

In this study, we used a combination of quantitative proteomics and bioinformatic prediction for identifying novel antioxidant peptides. Thirty-five peptides from potato, seaweed, microbial, and spinach proteins were investigated. Based on high DPPH radical scavenging activity (IC50 ≤ 16 mg/mL), metal chelation activity, isoelectric point, and high relative abundance in the parent protein sources, 11 peptides were selected. Lipid oxidation retardation was evaluated in 5% fish oil-in-water emulsions stabilized with Tween 20, where emulsion physical stability was unaffected by peptide addition. The secondary structure of selected peptides was similar in the aqueous solution and emulsions, as confirmed by synchrotron radiation circular dichroism spectroscopy. The emulsions containing the selected peptides had lower levels of hydroperoxides and volatile compounds during storage compared to the control (without peptide). This study contributes to elucidating the effect of antioxidant peptides in emulsions and demonstrates the ability of quantitative proteomics and bioinformatics prediction to identify peptides with strong antioxidant properties.

The Protein-Templated Synthesis of Enzyme-Generated Aptamers.

Inspired by the chemical synthesis of molecularly imprinted polymers, we demonstrated for the first time, the protein-target mediated synthesis of enzyme-generated aptamers (EGAs). We prepared pre-polymerisation mixtures containing different ratios of nucleotides, an initiator sequence and protein template and incubated each mixture with terminal deoxynucleotidyl transferase (TdT). Upon purification and rebinding of the EGAs against the target, we observed an enhancement in binding of templated-EGAs towards the target compared to a non-templated control. These results demonstrate the presence of two primary mechanisms for the formation of EGAs, namely, the binding of random sequences to the target as observed in systematic evolution of ligands by exponential enrichment (SELEX) and the dynamic competition between TdT enzyme and the target protein for binding of EGAs during synthesis. The latter mechanism serves to increase the stringency of EGA-based screening and represents a new way to develop aptamers that relies on rational design.

Variability analysis of LC-MS experimental factors and their impact on machine learning.

BACKGROUND: Machine learning (ML) technologies, especially deep learning (DL), have gained increasing attention in predictive mass spectrometry (MS) for enhancing the data-processing pipeline from raw data analysis to end-user predictions and rescoring. ML models need large-scale datasets for training and repurposing, which can be obtained from a range of public data repositories. However, applying ML to public MS datasets on larger scales is challenging, as they vary widely in terms of data acquisition methods, biological systems, and experimental designs. RESULTS: We aim to facilitate ML efforts in MS data by conducting a systematic analysis of the potential sources of variability in public MS repositories. We also examine how these factors affect ML performance and perform a comprehensive transfer learning to evaluate the benefits of current best practice methods in the field for transfer learning. CONCLUSIONS: Our findings show significantly higher levels of homogeneity within a project than between projects, which indicates that it is important to construct datasets most closely resembling future test cases, as transferability is severely limited for unseen datasets. We also found that transfer learning, although it did increase model performance, did not increase model performance compared to a non-pretrained model.

T Cell Epitope Prediction and Its Application to Immunotherapy.

T cells play a crucial role in controlling and driving the immune response with their ability to discriminate peptides derived from healthy as well as pathogenic proteins. In this review, we focus on the currently available computational tools for epitope prediction, with a particular focus on tools aimed at identifying neoepitopes, i.e. cancer-specific peptides and their potential for use in immunotherapy for cancer treatment. This review will cover how these tools work, what kind of data they use, as well as pros and cons in their respective applications.

Strategy to prevent epitope masking in CAR.CD19+ B-cell leukemia blasts.

Chimeric antigen receptor T-cells (CAR T-cells) for the treatment of relapsing/refractory B-cell precursor acute lymphoblastic leukemia have led to exciting clinical results. However, CAR T-cell approaches revealed a potential risk of CD19-/CAR+ leukemic relapse due to inadvertent transduction of leukemia cells. BACKGROUND: METHODS: We evaluated the impact of a high percentage of leukemia blast contamination in patient-derived starting material (SM) on CAR T-cell drug product (DP) manufacturing. In vitro as well as in vivo models were employed to identify characteristics of the construct associated with better profile of safety in case of inadvertent B-cell leukemia transduction during CAR T-cell manufacturing. RESULTS: The presence of large amounts of CD19+ cells in SM did not affect the transduction level of DPs, as well as the CAR T-cell rate of expansion at the end of standard production of 14 days. DPs were deeply characterized by flow cytometry and molecular biology for Ig-rearrangements, showing that the level of B-cell contamination in DPs did not correlate with the percentage of CD19+ cells in SM, in the studied patient cohort. Moreover, we investigated whether CAR design may affect the control of CAR+ leukemia cells. We provided evidences that CAR.CD19 short linker (SL) prevents complete epitope masking in CD19+CAR+ leukemia cells and we demonstrated in vitro and in vivo that CD19 +CAR(SL)+leukemic cells are killed by CAR.CD19 T-cells. CONCLUSIONS: Taken together, these data suggest that a VL-VH SL may result in a safe CAR-T product, even when manufacturing starts from biological materials characterized by heavy contamination of leukemia blasts.

Emulsifier peptides derived from seaweed, methanotrophic bacteria, and potato proteins identified by quantitative proteomics and bioinformatics.

Global focus on sustainability has accelerated research into alternative non-animal sources of food protein and functional food ingredients. Amphiphilic peptides represent a class of promising biomolecules to replace chemical emulsifiers in food emulsions. In contrast to traditional trial-and-error enzymatic hydrolysis, this study utilizes a bottom-up approach combining quantitative proteomics, bioinformatics prediction, and functional validation to identify novel emulsifier peptides from seaweed, methanotrophic bacteria, and potatoes. In vitro functional validation reveal that all protein sources contained embedded novel emulsifier peptides comparable to or better than sodium caseinate (CAS). Thus, peptides efficiently reduced oil-water interfacial tension and generated physically stable emulsions with higher net zeta potential and smaller droplet sizes than CAS. In silico structure modelling provided further insight on peptide structure and the link to emulsifying potential. This study clearly demonstrates the potential and broad applicability of the bottom-up approach for identification of abundant and potent emulsifier peptides.

Data mining patented antibody sequences.

The patent literature should reflect the past 30 years of engineering efforts directed toward developing monoclonal antibody therapeutics. Such information is potentially valuable for rational antibody design. Patents, however, are designed not to convey scientific knowledge, but to provide legal protection. It is not obvious whether antibody information from patent documents, such as antibody sequences, is useful in conveying engineering know-how, rather than as a legal reference only. To assess the utility of patent data for therapeutic antibody engineering, we quantified the amount of antibody sequences in patents destined for medicinal purposes and how well they reflect the primary sequences of therapeutic antibodies in clinical use. We identified 16,526 patent families covering major jurisdictions (e.g., US Patent and Trademark Office (USPTO) and World Intellectual Property Organization) that contained antibody sequences. These families held 245,109 unique antibody chains (135,397 heavy chains and 109,712 light chains) that we compiled in our Patented Antibody Database (PAD, http://naturalantibody.com/pad). We find that antibodies make up a non-trivial proportion of all patent amino acid sequence depositions (e.g., 11% of USPTO Full Text database). Our analysis of the 16,526 families demonstrates that the volume of patent documents with antibody sequences is growing, with the majority of documents classified as containing antibodies for medicinal purposes. We further studied the 245,109 antibody chains from patent literature to reveal that they very well reflect the primary sequences of antibody therapeutics in clinical use. This suggests that the patent literature could serve as a reference for previous engineering efforts to improve rational antibody design.

Terminal deoxynucleotidyl transferase-mediated formation of protein binding polynucleotides.

Terminal deoxynucleotidyl transferase (TdT) enzyme plays an integral part in the V(D)J recombination, allowing for the huge diversity in expression of immunoglobulins and T-cell receptors within lymphocytes, through their unique ability to incorporate single nucleotides into oligonucleotides without the need of a template. The role played by TdT in lymphocytes precursors found in early vertebrates is not known. In this paper, we demonstrated a new screening method that utilises TdT to form libraries of variable sized (vsDNA) libraries of polynucleotides that displayed binding towards protein targets. The extent of binding and size distribution of each vsDNA library towards their respective protein target can be controlled through the alteration of different reaction conditions such as time of reaction, nucleotide ratio and initiator concentration raising the possibility for the rational design of aptamers prior to screening. The new approach, allows for the screening of aptamers based on size as well as sequence in a single round, which minimises PCR bias. We converted the protein bound sequences to dsDNA using rapid amplification of variable ends assays (RAVE) and sequenced them using next generation sequencing. The resultant aptamers demonstrated low nanomolar binding and high selectivity towards their respective targets.

Selection and Modelling of a New Single-Domain Intrabody Against TDP-43.

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder associated to deteriorating motor and cognitive functions, and short survival. The disease is caused by neuronal death which results in progressive muscle wasting and weakness, ultimately leading to lethal respiratory failure. The misbehaviour of a specific protein, TDP-43, which aggregates and becomes toxic in ALS patient's neurons, is supposed to be one of the causes. TDP-43 is a DNA/RNA-binding protein involved in several functions related to nucleic acid metabolism. Sequestration of TDP-43 aggregates is a possible therapeutic strategy that could alleviate or block pathology. Here, we describe the selection and characterization of a new intracellular antibody (intrabody) against TDP-43 from a llama nanobody library. The structure of the selected intrabody was predicted in silico and the model was used to suggest mutations that enabled to improve its expression yield, facilitating its experimental validation. We showed how coupling experimental methodologies with in silico design may allow us to obtain an antibody able to recognize the RNA binding regions of TDP-43. Our findings illustrate a strategy for the mitigation of TDP-43 proteinopathy in ALS and provide a potential new tool for diagnostics.

AnOxPePred: using deep learning for the prediction of antioxidative properties of peptides.

Dietary antioxidants are an important preservative in food and have been suggested to help in disease prevention. With consumer demands for less synthetic and safer additives in food products, the food industry is searching for antioxidants that can be marketed as natural. Peptides derived from natural proteins show promise, as they are generally regarded as safe and potentially contain other beneficial bioactivities. Antioxidative peptides are usually obtained by testing various peptides derived from hydrolysis of proteins by a selection of proteases. This slow and cumbersome trial-and-error approach to identify antioxidative peptides has increased interest in developing computational approaches for prediction of antioxidant activity and thereby reduce laboratory work. A few antioxidant predictors exist, however, no tool predicting the antioxidative properties of peptides is, to the best of our knowledge, currently available as a web-server. We here present the AnOxPePred tool and web-server ( http://services.bioinformatics.dtu.dk/service.php?AnOxPePred-1.0 ) that uses deep learning to predict the antioxidant properties of peptides. Our model was trained on a curated dataset consisting of experimentally-tested antioxidant and non-antioxidant peptides. For a variety of metrics our method displays a prediction performance better than a k-NN sequence identity-based approach. Furthermore, the developed tool will be a good benchmark for future predictors of antioxidant peptides.

Biofunctionality of Enzymatically Derived Peptides from Codfish (Gadus morhua) Frame: Bulk In Vitro Properties, Quantitative Proteomics, and Bioinformatic Prediction.

Protein hydrolysates show great promise as bioactive food and feed ingredients and for valorization of side-streams from e.g., the fish processing industry. We present a novel approach for hydrolysate characterization that utilizes proteomics data for calculation of weighted mean peptide properties (length, molecular weight, and charge) and peptide-level abundance estimation. Using a novel bioinformatic approach for subsequent prediction of biofunctional properties of identified peptides, we are able to provide an unprecedented, in-depth characterization. The study further characterizes bulk emulsifying, foaming, and in vitro antioxidative properties of enzymatic hydrolysates derived from cod frame by application of Alcalase and Neutrase, individually and sequentially, as well as the influence of heat pre-treatment. All hydrolysates displayed comparable or higher emulsifying activity and stability than sodium caseinate. Heat-treatment significantly increased stability but showed a negative effect on the activity and degree of hydrolysis. Lower degrees of hydrolysis resulted in significantly higher chelating activity, while the opposite was observed for radical scavenging activity. Combining peptide abundance with bioinformatic prediction, we identified several peptides that are likely linked to the observed differences in bulk emulsifying properties. The study highlights the prospects of applying proteomics and bioinformatics for hydrolysate characterization and in food protein science.