Search

The ultimate goal of protein design is to introduce new biological activity. We propose a computational approach for designing functional antibodies by focusing on functional epitopes, integrating large-scale statistical analysis with multiple structural models. Machine learning is used to analyze these models and predict specific residue-residue contacts. We use this approach to design a functional antibody to counter the proinflammatory effect of the cytokine interleukin-17A (IL-17A). X-ray crystallography confirms that the designed antibody binds the targeted epitope and the interaction is mediated by the designed contacts. Cell-based assays confirm that the antibody is functional. Importantly, this approach does not rely on a high-quality 3D model of the designed complex or even a solved structure of the target. As demonstrated here, this approach can be used to design biologically active antibodies, removing some of the main hurdles in antibody design and in drug discovery., (Copyright © 2018 The Authors. Published by Elsevier Inc. All rights reserved.)

The development of accurate tools for predicting B-cell epitopes is important but difficult. Traditional methods have examined which regions in an antigen are likely binding sites of an antibody. However, it is becoming increasingly clear that most antigen surface residues will be able to bind one or more of the myriad of possible antibodies. In recent years, new approaches have emerged for predicting an epitope for a specific antibody, utilizing information encoded in antibody sequence or structure. Applying such antibody-specific predictions to groups of antibodies in combination with easily obtainable experimental data improves the performance of epitope predictions. We expect that further advances of such tools will be possible with the integration of immunoglobulin repertoire sequencing data., (Copyright © 2015 Elsevier B.V. All rights reserved.)

Spatiotemporal coordination is a critical factor in biological processes. Some hubs in protein-protein interaction networks tend to be co-expressed and co-localized with their partners more strongly than others, a difference which is arguably related to functional differences between the hubs. Based on numerous analyses of yeast hubs, it has been suggested that differences in co-expression and co-localization are reflected in the structural and molecular characteristics of the hubs. We hypothesized that if indeed differences in co-expression and co-localization are encoded in the molecular characteristics of the protein, it may be possible to predict the tendency for co-expression and co-localization of human hubs based on features learned from systematically characterized yeast hubs. Thus, we trained a prediction algorithm on hubs from yeast that were classified as either strongly or weakly co-expressed and co-localized with their partners, and applied the trained model to 800 human hub proteins. We found that the algorithm significantly distinguishes between human hubs that are co-expressed and co-localized with their partners and hubs that are not. The prediction is based on sequence derived features such as "stickiness", i.e. the existence of multiple putative binding sites that enable multiple simultaneous interactions, "plasticity", i.e. the existence of predicted structural disorder which conjecturally allows for multiple consecutive interactions with the same binding site and predicted subcellular localization. These results suggest that spatiotemporal dynamics is encoded, at least in part, in the amino acid sequence of the protein and that this encoding is similar in yeast and in human.

General protein-protein interfaces are known to be enriched, compared with other surface patches, with amino acids that can form stabilizing interactions. However, several studies reported that there are hardly any differences between the amino acid composition of B-cell epitopes and that of antigen surface residues. If the amino acid composition of epitopes is indistinguishable from other surface patches, how do antibodies (Abs) identify epitopes? Here, we analyze the antigen binding regions (ABRs, roughly corresponding to the complementarity determining regions) and the epitopes in a non-redundant set of all known Ab-antigen complexes. We find that the ABRs differ significantly from each other in their amino acid composition and length. Analysis of the energetic contribution of each ABR to antigen binding reveals that, while H3 often plays a key role in antigen binding, in many antibodies other ABRs are more important. Moreover, each ABR has a distinct propensity to bind different amino acids on the antigen. The combined binding preferences of the ABRs yield a total preference to amino acids with a composition that is virtually identical to that of surface residues. These results suggest that antibodies evolved to recognize protein surfaces. They may help in improving Ab engineering and B-cell epitope prediction.

Every entirely sequenced genome reveals 100 s to 1000 s of protein sequences for which the only annotation available is 'hypothetical protein'. Thus, in the human genome and in the genomes of pathogenic agents there could be 1000 s of potential, unexplored drug targets. Computational prediction of protein function can play a role in studying these targets. We shall review the challenges, research approaches and recently developed tools in the field of computational function-prediction and we will discuss the ways these issues can change the process of drug discovery.

Most methods annotating protein function utilise sequence homology to proteins of experimentally known function. Such a homology-based annotation transfer is problematic and limited in scope. Therefore, computational biologists have begun to develop ab initio methods that predict aspects of function, including subcellular localization, post-translational modifications, functional type and protein-protein interactions. For the first two cases, the most accurate approaches rely on identifying short signalling motifs, while the most general methods utilise tools of artificial intelligence. An outstanding new method predicts classes of cellular function directly from sequence. Similarly, promising methods have been developed predicting protein-protein interaction partners at acceptable levels of accuracy for some pairs in entire proteomes. No matter how difficult the task, successes over the last few years have clearly paved the way for ab initio prediction of protein function.

Protein-protein interactions are facilitated by a myriad of residue-residue contacts on the interacting proteins. Identifying the site of interaction in the protein is a key for deciphering its functional mechanisms, and is crucial for drug development. Many studies indicate that the compositions of contacting residues are unique. Here, we describe a neural network that identifies protein-protein interfaces from sequence. For the most strongly predicted sites (in 34 of 333 proteins), 94% of the predictions were confirmed experimentally. When 70% of our predictions were right, we correctly predicted at least one interaction site in 20% of the complexes (66/333). These results indicate that the prediction of some interaction sites from sequence alone is possible. Incorporating evolutionary and predicted structural information may improve our method. However, even at this early stage, our tool might already assist wet-lab biology.