Customized versus universal scoring functions: application to class I MHC-peptide binding free energy predictions

A minigene DNA vaccine encoding peptide epitopes derived from Galectin-1 has protective antitumoral effects in a model of neuroblastoma

We recently identified Galectin-1 (Gal-1), a β-galactoside-binding lectin, as a novel immune regulator in neuroblastoma (NB). Here, we characterized the tolerogenic function of Gal-1 within the CD8⁺ T cell compartment and further evaluated its relevance as an antigen for effective DNA vaccination against NB in a mouse model. NB cells with Gal-1 knockdown (NXS-2L) exhibited significantly reduced tumor growth compared to NXS-2 NB cells. Administration of anti-CD8 antibodies prevented this antitumor effect, with primary tumor growth comparable to that from Gal-1 (G1)-sufficient NB cells. Peptide epitope screening with online databases and in silico docking experiments predicted the sequences "FDQADLTI" (#1), "GDFKIKCV" (#2), and "AHGDANTI" (#3) to have superior H2-KK binding affinities and "KFPNRLNM" (#4), "DGDFKIKCV" (#5), and "LGKDSNNL" (#6) to have superior H2-DD binding affinities. Minigenes encoding G1-KK (#1-#2-#3), G1-DD (#4-#5-#6) and the triplet with the highest affinity, G1-H (#1-#2-#4), were generated and cloned into a ubiquitin-containing plasmid (pU). Mice receiving pU-G1-KK or pU-G-1H presented a reduction in the s.c. tumor volume and weight of up to 80% compared to control mice; this reduction was associated with increased cytotoxicity of isolated splenocytes from vaccinated animals. Vaccination with pUG1-DD showed a lower capability to suppress primary tumor progression. In conclusion, Gal-1 expression by NB negatively regulates CD8⁺ T cells. Vaccination with DNA plasmids encoding Gal-1 epitopes overcomes immune escape, enhances CD8⁺ T cell-dependent immunity and displays effective antitumor activity against NB.

New machine learning and physics-based scoring functions for drug discovery

Scoring functions are essential for modern in silico drug discovery. However, the accurate prediction of binding affinity by scoring functions remains a challenging task. The performance of scoring functions is very heterogeneous across different target classes. Scoring functions based on precise physics-based descriptors better representing protein–ligand recognition process are strongly needed. We developed a set of new empirical scoring functions, named DockTScore, by explicitly accounting for physics-based terms combined with machine learning. Target-specific scoring functions were developed for two important drug targets, proteases and protein–protein interactions, representing an original class of molecules for drug discovery. Multiple linear regression (MLR), support vector machine and random forest algorithms were employed to derive general and target-specific scoring functions involving optimized MMFF94S force-field terms, solvation and lipophilic interactions terms, and an improved term accounting for ligand torsional entropy contribution to ligand binding. DockTScore scoring functions demonstrated to be competitive with the current best-evaluated scoring functions in terms of binding energy prediction and ranking on four DUD-E datasets and will be useful for in silico drug design for diverse proteins as well as for specific targets such as proteases and protein–protein interactions. Currently, the MLR DockTScore is available at www.dockthor.lncc.br.

Large-Scale Structure-Based Prediction of Stable Peptide Binding to Class I HLAs Using Random Forests

Prediction of stable peptide binding to Class I HLAs is an important component for designing immunotherapies. While the best performing predictors are based on machine learning algorithms trained on peptide-HLA (pHLA) sequences, the use of structure for training predictors deserves further exploration. Given enough pHLA structures, a predictor based on the residue-residue interactions found in these structures has the potential to generalize for alleles with little or no experimental data. We have previously developed APE-Gen, a modeling approach able to produce pHLA structures in a scalable manner. In this work we use APE-Gen to model over 150,000 pHLA structures, the largest dataset of its kind, which were used to train a structure-based pan-allele model. We extract simple, homogenous features based on residue-residue distances between peptide and HLA, and build a random forest model for predicting stable pHLA binding. Our model achieves competitive AUROC values on leave-one-allele-out validation tests using significantly less data when compared to popular sequence-based methods. Additionally, our model offers an interpretation analysis that can reveal how the model composes the features to arrive at any given prediction. This interpretation analysis can be used to check if the model is in line with chemical intuition, and we showcase particular examples. Our work is a significant step toward using structure to achieve generalizable and more interpretable prediction for stable pHLA binding.

Empirical Scoring Functions for Structure-Based Virtual Screening: Applications, Critical Aspects, and Challenges

Structure-based virtual screening (VS) is a widely used approach that employs the knowledge of the three-dimensional structure of the target of interest in the design of new lead compounds from large-scale molecular docking experiments. Through the prediction of the binding mode and affinity of a small molecule within the binding site of the target of interest, it is possible to understand important properties related to the binding process. Empirical scoring functions are widely used for pose and affinity prediction. Although pose prediction is performed with satisfactory accuracy, the correct prediction of binding affinity is still a challenging task and crucial for the success of structure-based VS experiments. There are several efforts in distinct fronts to develop even more sophisticated and accurate models for filtering and ranking large libraries of compounds. This paper will cover some recent successful applications and methodological advances, including strategies to explore the ligand entropy and solvent effects, training with sophisticated machine-learning techniques, and the use of quantum mechanics. Particular emphasis will be given to the discussion of critical aspects and further directions for the development of more accurate empirical scoring functions.

Prediction of peptide binding to a major histocompatibility complex class I molecule based on docking simulation

Binding between major histocompatibility complex (MHC) class I molecules and immunogenic epitopes is one of the most important processes for cell-mediated immunity. Consequently, computational prediction of amino acid sequences of MHC class I binding peptides from a given sequence may lead to important biomedical advances. In this study, an efficient structure-based method for predicting peptide binding to MHC class I molecules was developed, in which the binding free energy of the peptide was evaluated by two individual docking simulations. An original penalty function and restriction of degrees of freedom were determined by analysis of 361 published X-ray structures of the complex and were then introduced into the docking simulations. To validate the method, calculations using a 50-amino acid sequence as a prediction target were performed. In 27 calculations, the binding free energy of the known peptide was within the top 5 of 166 peptides generated from the 50-amino acid sequence. Finally, demonstrative calculations using a whole sequence of a protein as a prediction target were performed. These data clearly demonstrate high potential of this method for predicting peptide binding to MHC class I molecules.

Machine Learning Methods for Predicting HLA-Peptide Binding Activity

As major histocompatibility complexes in humans, the human leukocyte antigens (HLAs) have important functions to present antigen peptides onto T-cell receptors for immunological recognition and responses. Interpreting and predicting HLA–peptide binding are important to study T-cell epitopes, immune reactions, and the mechanisms of adverse drug reactions. We review different types of machine learning methods and tools that have been used for HLA–peptide binding prediction. We also summarize the descriptors based on which the HLA–peptide binding prediction models have been constructed and discuss the limitation and challenges of the current methods. Lastly, we give a future perspective on the HLA–peptide binding prediction method based on network analysis.

Receptor-ligand molecular docking

Docking methodology aims to predict the experimental binding modes and affinities of small molecules within the binding site of particular receptor targets and is currently used as a standard computational tool in drug design for lead compound optimisation and in virtual screening studies to find novel biologically active molecules. The basic tools of a docking methodology include a search algorithm and an energy scoring function for generating and evaluating ligand poses. In this review, we present the search algorithms and scoring functions most commonly used in current molecular docking methods that focus on protein-ligand applications. We summarise the main topics and recent computational and methodological advances in protein-ligand docking. Protein flexibility, multiple ligand binding modes and the free-energy landscape profile for binding affinity prediction are important and interconnected challenges to be overcome by further methodological developments in the docking field.

Computational Resources for MHC Ligand Identification

Advances in the high-throughput determination of functional modulators of major histocompatibility complex (MHC) and improved computational predictions of MHC ligands have rendered the rational design of immunomodulatory peptides feasible. Proteome-derived peptides and 'reverse vaccinology' by computational means will play a driving role in future vaccine design. Here we review the molecular mechanisms of the MHC mediated immune response, present the computational approaches that have emerged in this area of biotechnology, and provide an overview of publicly available computational resources for predicting and designing new peptidic MHC ligands.

T-cell epitope vaccine design by immunoinformatics

Vaccination is generally considered to be the most effective method of preventing infectious diseases. All vaccinations work by presenting a foreign antigen to the immune system in order to evoke an immune response. The active agent of a vaccine may be intact but inactivated ('attenuated') forms of the causative pathogens (bacteria or viruses), or purified components of the pathogen that have been found to be highly immunogenic. The increased understanding of antigen recognition at molecular level has resulted in the development of rationally designed peptide vaccines. The concept of peptide vaccines is based on identification and chemical synthesis of B-cell and T-cell epitopes which are immunodominant and can induce specific immune responses. The accelerating growth of bioinformatics techniques and applications along with the substantial amount of experimental data has given rise to a new field, called immunoinformatics. Immunoinformatics is a branch of bioinformatics dealing with in silico analysis and modelling of immunological data and problems. Different sequence- and structure-based immunoinformatics methods are reviewed in the paper.

T-cell epitope prediction and immune complex simulation using molecular dynamics: state of the art and persisting challenges

Atomistic Molecular Dynamics provides powerful and flexible tools for the prediction and analysis of molecular and macromolecular systems. Specifically, it provides a means by which we can measure theoretically that which cannot be measured experimentally: the dynamic time-evolution of complex systems comprising atoms and molecules. It is particularly suitable for the simulation and analysis of the otherwise inaccessible details of MHC-peptide interaction and, on a larger scale, the simulation of the immune synapse. Progress has been relatively tentative yet the emergence of truly high-performance computing and the development of coarse-grained simulation now offers us the hope of accurately predicting thermodynamic parameters and of simulating not merely a handful of proteins but larger, longer simulations comprising thousands of protein molecules and the cellular scale structures they form. We exemplify this within the context of immunoinformatics.

In silico quantitative prediction of peptides binding affinity to human MHC molecule: an intuitive quantitative structure-activity relationship approach

In this paper, we have handpicked 23 kinds of electronic properties, 37 kinds of steric properties, 54 kinds of hydrophobic properties and 5 kinds of hydrogen bond properties from thousands of amino acid structural and property parameters. Principal component analysis (PCA) was applied on these parameters and thus ten score vectors involving significant nonbonding properties of 20 coded amino acids were yielded, called the divided physicochemical property scores (DPPS) of amino acids. The DPPS descriptor was then used to characterize the structures of 152 HLA-A*0201-restricted CTL epitopes, and significant variables being responsible for the binding affinities were selected by genetic algorithm, and a quantitative structure-activity relationship (QSAR) model by partial least square was established to predict the peptide-HLA-A*0201 molecule interactions. Statistical analysis on the resulted DPPS-based QSAR models were consistent well with experimental exhibits and molecular graphics display. Diversified properties of the different residues in binding peptides may contribute remarkable effect to the interactions between the HLA-A*0201 molecule and its peptide ligands. Particularly, hydrophobicity and hydrogen bond of anchor residues of peptides may have a significant contribution to the interactions. The results showed that DPPS can well represent the structural characteristics of the antigenic peptides and is a promising approach to predict the affinities of peptide binding to HLA-A*0201 in a efficient and intuitive way. We expect that this physical-principle based method can be applied to other protein-peptide interactions as well.

Modeling the adaptive immune system: predictions and simulations

MOTIVATION

Immunological bioinformatics methods are applicable to a broad range of scientific areas. The specifics of how and where they might be implemented have recently been reviewed in the literature. However, the background and concerns for selecting between the different available methods have so far not been adequately covered.

SUMMARY

Before using predictions systems, it is necessary to not only understand how the methods are constructed but also their strength and limitations. The prediction systems in humoral epitope discovery are still in their infancy, but have reached a reasonable level of predictive strength. In cellular immunology, MHC class I binding predictions are now very strong and cover most of the known HLA specificities. These systems work well for epitope discovery, and predictions of the MHC class I pathway have been further improved by integration with state-of-the-art prediction tools for proteasomal cleavage and TAP binding. By comparison, class II MHC binding predictions have not developed to a comparable accuracy level, but new tools have emerged that deliver significantly improved predictions not only in terms of accuracy, but also in MHC specificity coverage. Simulation systems and mathematical modeling are also now beginning to reach a level where these methods will be able to answer more complex immunological questions.

HLA-peptide binding prediction using structural and modeling principles

Short peptides binding to specific human leukocyte antigen (HLA) alleles elicit immune response. These candidate peptides have potential utility in peptide vaccine design and development. The binding of peptides to allele-specific HLA molecule is estimated using competitive binding assay and biochemical binding constants. Application of this method for proteome-wide screening in parasites, viruses, and virulent bacterial strains is laborious and expensive. However, short listing of candidate peptides using prediction approaches have been realized lately. Prediction of peptide binding to HLA alleles using structural and modeling principles has gained momentum in recent years. Here, we discuss the current status of such prediction.

Nonlinear predictive modeling of MHC class II-peptide binding using Bayesian neural networks

Methods for predicting the binding affinity of peptides to the MHC have become more sophisticated in the past 5-10 years. It is possible to use computational quantitative structure-activity methods to build models of peptide affinity that are truly predictive. Two of the most useful methods for building models are Bayesian regularized neural networks for continuous or discrete (categorical) data and support vector machines (SVMs) for discrete data. We illustrate the application of Bayesian regularized neural networks to modeling MHC class II-binding affinity of peptides. Training data comprised sequences and binding data for nonamer (nine amino acid) peptides. Peptides were characterized by mathematical representations of several types. Independent test data comprised sequences and binding data for peptides of length < or = 25. We also internally validated the models by using 30% of the data in an internal test set. We obtained robust models, with near-identical statistics for multiple training runs. We determined how predictive our models were using statistical tests and area under the receiver operating characteristic (ROC) graphs (A(ROC)). Some mathematical representations of the peptides were more efficient than others and were able to generalize to unknown peptides outside of the training space. Bayesian neural networks are robust, efficient "universal approximators" that are well able to tackle the difficult problem of correctly predicting the MHC class II-binding activities of a majority of the test set peptides.

Predictive Bayesian neural network models of MHC class II peptide binding

We used Bayesian regularized neural networks to model data on the MHC class II-binding affinity of peptides. Training data consisted of sequences and binding data for nonamer (nine amino acid) peptides. Independent test data consisted of sequences and binding data for peptides of length </=25. We assumed that MHC class II-binding activity of peptides depends only on the highest ranked embedded nonamer and that reverse sequences of active nonamers are inactive. We also internally validated the models by using 30% of the training data in an internal test set. We obtained robust models, with near identical statistics for multiple training runs. We determined how predictive our models were using statistical tests and area under the Receiver Operating Characteristic (ROC) graphs (A(ROC)). Most models gave training A(ROC) values close to 1.0 and test set A(ROC) values >0.8. We also used both amino acid indicator variables (bin20) and property-based descriptors to generate models for MHC class II-binding of peptides. The property-based descriptors were more parsimonious than the indicator variable descriptors, making them applicable to larger peptides, and their design makes them able to generalize to unknown peptides outside of the training space. None of the external test data sets contained any of the nonamer sequences in the training sets. Consequently, the models attempted to predict the activity of truly unknown peptides not encountered in the training sets. Our models were well able to tackle the difficult problem of correctly predicting the MHC class II-binding activities of a majority of the test set peptides. Exceptions to the assumption that nonamer motif activities were invariant to the peptide in which they were embedded, together with the limited coverage of the test data, and the fuzziness of the classification procedure, are likely explanations for some misclassifications.

Structural mining: self-consistent design on flexible protein-peptide docking and transferable binding affinity potential

A flexible protein-peptide docking method has been designed to consider not only ligand flexibility but also the flexibility of the protein. The method is based on a Monte Carlo annealing process. Simulations with a distance root-mean-square (dRMS) virtual energy function revealed that the flexibility of protein side chains was as important as ligand flexibility for successful protein-peptide docking. On the basis of mean field theory, a transferable potential was designed to evaluate distance-dependent protein-ligand interactions and atomic solvation energies. The potential parameters were developed using a self-consistent process based on only 10 known complex structures. The effectiveness of each intermediate potential was judged on the basis of a Z score, approximating the gap between the energy of the native complex and the average energy of a decoy set. The Z score was determined using experimentally determined native structures and decoys generated by docking with the intermediate potentials. Using 6600 generated decoys and the Z score optimization criterion proposed in this work, the developed potential yielded an acceptable correlation of R(2) = 0.77, with binding free energies determined for known MHC I complexes (Class I Major Histocompatibility protein HLA-A(*)0201) which were not present in the training set. Test docking on 25 complexes further revealed a significant correlation between energy and dRMS, important for identifying native-like conformations. The near-native structures always belonged to one of the conformational classes with lower predicted binding energy. The lowest energy docked conformations are generally associated with near-native conformations, less than 3.0 Angstrom dRMS (and in many cases less than 1.0 Angstrom) from the experimentally determined structures.

Towards in silico prediction of immunogenic epitopes

As torrents of new data now emerge from microbial genomics, bioinformatic prediction of immunogenic epitopes remains challenging but vital. In silico methods often produce paradoxically inconsistent results: good prediction rates on certain test sets but not others. The inherent complexity of immune presentation and recognition processes complicates epitope prediction. Two encouraging developments - data driven artificial intelligence sequence-based methods for epitope prediction and molecular modeling methods based on three-dimensional protein structures - offer hope for the future.

JenPep: a novel computational information resource for immunobiology and vaccinology

JenPep is a relational database containing a compendium of thermodynamic binding data for the interaction of peptides with a range of important immunological molecules: the major histocompatibility complex, TAP transporter, and T cell receptor. The database also includes annotated lists of B cell and T cell epitopes. Version 2.0 of the database is implemented in a bespoke postgreSQL database system and is fully searchable online via a perl/HTML interface (URL: http://www.jenner.ac.uk/JenPep).

Proteomics in Vaccinology and Immunobiology: An Informatics Perspective of the Immunone

The postgenomic era, as manifest, inter alia, by proteomics, offers unparalleled opportunities for the efficient discovery of safe, efficacious, and novel subunit vaccines targeting a tranche of modern major diseases. A negative corollary of this opportunity is the risk of becoming overwhelmed by this embarrassment of riches. Informatics techniques, working to address issues of both data management and through prediction to shortcut the experimental process, can be of enormous benefit in leveraging the proteomic revolution. In this disquisition, we evaluate proteomic approaches to the discovery of subunit vaccines, focussing on viral, bacterial, fungal, and parasite systems. We also adumbrate the impact that proteomic analysis of host-pathogen interactions can have. Finally, we review relevant methods to the prediction of immunome, with special emphasis on quantitative methods, and the subcellular localization of proteins within bacteria.

Molecular mimicry of an HLA-B27-derived ligand of arthritis-linked subtypes with chlamydial proteins

HLA-B27 is strongly associated with spondyloarthropathies, including ankylosing spondylitis and reactive arthritis. The latter disease is triggered by various Gram-negative bacteria. A dodecamer derived from the intracytoplasmic tail of HLA-B27 was a natural ligand of three disease-associated subtypes (B*2702, B*2704, and B*2705) but not of two (B*2706 and B*2709), weakly or not associated to spondyloarthropathy. This peptide was strikingly homologous to protein sequences from arthritogenic bacteria, particularly to a region of the DNA primase from Chlamydia trachomatis. A synthetic peptide with this bacterial sequence bound in vitro disease-associated subtypes equally as the natural B27-derived ligand. The chlamydial peptide was generated by the 20 S proteasome from a synthetic 28-mer with the sequence of the corresponding region of the bacterial DNA primase. Molecular modeling suggested that the B27-derived and chlamydial peptides adopt very similar conformations in complex with B*2705. The results demonstrate that an HLA-B27-derived peptide mimicking arthritogenic bacterial sequences is a natural ligand of disease-associated HLA-B27 subtypes and suggest that the homologous chlamydial peptide might be presented by HLA-B27 on Chlamydia-infected cells.

Physicochemical explanation of peptide binding to HLA-A*0201 major histocompatibility complex: a three-dimensional quantitative structure-activity relationship study

A three-dimensional quantitative structure-activity relationship method for the prediction of peptide binding affinities to the MHC class I molecule HLA-A*0201 was developed by applying the CoMSIA technique on a set of 266 peptides. To increase the self consistency of the initial CoMSIA model, the poorly predicted peptides were excluded from the training set in a stepwise manner and then included in the study as a test set. The final model, based on 236 peptides and considering the steric, electrostatic, hydrophobic, hydrogen bond donor, and hydrogen bond acceptor fields, had q2 = 0.683 and r2 = 0.891. The stability of this model was proven by cross-validations in two and five groups and by a bootstrap analysis of the non-cross-validated model. The residuals between the experimental pIC50 (-logIC50) values and those calculated by "leave-one-out" cross-validation were analyzed. According to the best model, 63.2% of the peptides were predicted with /residuals/ < or = 0.5 log unit; 29.3% with 1.0 < or = /residuals/ < 0.5; and 7.5% with /residuals/ > 1.0 log unit. The mean /residual/ value was 0.489. The coefficient contour maps identify the physicochemical property requirements at each position in the peptide molecule and suggest amino acid sequences for high-affinity binding to the HLA-A*0201 molecule.

Principles of docking: An overview of search algorithms and a guide to scoring functions

The docking field has come of age. The time is ripe to present the principles of docking, reviewing the current state of the field. Two reasons are largely responsible for the maturity of the computational docking area. First, the early optimism that the very presence of the "correct" native conformation within the list of predicted docked conformations signals a near solution to the docking problem, has been replaced by the stark realization of the extreme difficulty of the next scoring/ranking step. Second, in the last couple of years more realistic approaches to handling molecular flexibility in docking schemes have emerged. As in folding, these derive from concepts abstracted from statistical mechanics, namely, populations. Docking and folding are interrelated. From the purely physical standpoint, binding and folding are analogous processes, with similar underlying principles. Computationally, the tools developed for docking will be tremendously useful for folding. For large, multidomain proteins, domain docking is probably the only rational way, mimicking the hierarchical nature of protein folding. The complexity of the problem is huge. Here we divide the computational docking problem into its two separate components. As in folding, solving the docking problem involves efficient search (and matching) algorithms, which cover the relevant conformational space, and selective scoring functions, which are both efficient and effectively discriminate between native and non-native solutions. It is universally recognized that docking of drugs is immensely important. However, protein-protein docking is equally so, relating to recognition, cellular pathways, and macromolecular assemblies. Proteins function when they are bound to other molecules. Consequently, we present the review from both the computational and the biological points of view. Although large, it covers only partially the extensive body of literature, relating to small (drug) and to large protein-protein molecule docking, to rigid and to flexible. Unfortunately, when reviewing these, a major difficulty in assessing the results is the non-uniformity in the formats in which they are presented in the literature. Consequently, we further propose a way to rectify it here.

Recovery of known T-cell epitopes by computational scanning of a viral genome

A new computational method (EpiDock) is proposed for predicting peptide binding to class I MHC proteins, from the amino acid sequence of any protein of immunological interest. Starting from the primary structure of the target protein, individual three-dimensional structures of all possible MHC-peptide (8-, 9- and 10-mers) complexes are obtained by homology modelling. A free energy scoring function (Fresno) is then used to predict the absolute binding free energy of all possible peptides to the class I MHC restriction protein. Assuming that immunodominant epitopes are usually found among the top MHC binders, the method can thus be applied to predict the location of immunogenic peptides on the sequence of the protein target. When applied to the prediction of HLA-A*0201-restricted T-cell epitopes from the Hepatitis B virus, EpiDock was able to recover 92% of known high affinity binders and 80% of known epitopes within a filtered subset of all possible nonapeptides corresponding to about one tenth of the full theoretical list. The proposed method is fully automated and fast enough to scan a viral genome in less than an hour on a parallel computing architecture. As it requires very few starting experimental data, EpiDock can be used: (i) to predict potential T-cell epitopes from viral genomes (ii) to roughly predict still unknown peptide binding motifs for novel class I MHC alleles.