Abstract
Development of vaccines and therapeutic antibodies to deal with infectious and other diseases are the most perceptible scientific interventions that have had huge impact on public health including that in the current Covid-19 pandemic. From inactivation methodologies to reverse vaccinology, vaccine development strategies of 21st century have undergone several transformations and are moving towards rational design approaches. These developments are driven by data as the combinatorials involved in antigenic diversity of pathogens and immune repertoire of hosts are enormous. The computational prediction of epitopes is central to these developments and numerous B-cell epitope prediction methods developed over the years in the field of immunoinformatics have contributed enormously. Most of these methods predict epitopes that could potentially bind to an antibody regardless of its type and only a few account for antibody class specific epitope prediction. Recent studies have provided evidence of more than one class of antibodies being associated with a particular disease. Therefore, it is desirable to predict and prioritize ‘peptidome’ representing B-cell epitopes that can potentially bind to multiple classes of antibodies, as an open problem in immunoinformatics. To address this, AbCPE, a novel algorithm based on multi-label classification approach has been developed for prediction of antibody class(es) to which an epitope can potentially bind. The epitopes binding to one or more antibody classes (IgG, IgE, IgA and IgM) have been used as a knowledgebase to derive features for prediction. Multi-label algorithms, Binary Relevance and Label Powerset were applied along with Random Forest and AdaBoost. Classifier performance was assessed using evaluation measures like Hamming Loss, Precision, Recall and F1 score. The Binary Relevance model based on dipeptide composition, Random Forest and AdaBoost achieved the best results with Hamming Loss of 0.1121 and 0.1074 on training and test sets respectively. The results obtained by AbCPE are promising. To the best of our knowledge, this is the first multi-label method developed for prediction of antibody class(es) for sequential B-cell epitopes and is expected to bring a paradigm shift in the field of immunoinformatics and immunotherapeutic developments in synthetic biology. The AbCPE web server is available at http://bioinfo.unipune.ac.in/AbCPE/Home.html.
Highlights
Antibody-mediated immune response is characterized by generation of antibodies from activated B-cells which are targeted at specific pathogens or pathogenic molecules
For epitope classes that bind to more than one class of antibody, differences in percent amino acid composition can be seen for most of the residues with some explicit trends, like for amino acids glutamine (Q), proline (P) and valine (V) that are more predominantly present while histidine (H) and cysteine (C) are less common. These observations indicate that almost all the 20 amino acids show variation in their occurrence in epitopes binding to specific antibody class/es. This variation in amino acid composition can be utilized in terms of compositional features to design and develop the models and the algorithms for prediction of epitopes that bind to specific antibody class/es
The field of computational B-cell epitope prediction has progressed and evolved at a tremendous pace in recent years with availability of large number of methods which have accelerated the pace of rational design of vaccines
Summary
Antibody-mediated immune response is characterized by generation of antibodies (immunoglobulins) from activated B-cells which are targeted at specific pathogens or pathogenic molecules (antigens). Many computational methods have been developed for prediction of linear as well as conformational epitopes (Yao et al, 2013; Sanchez-Trincado et al, 2017) and some of these have been made available on IEDB portal. Numerous epitope prediction methods based on machine learning algorithms have been developed recently, which utilize variety of features derived from sequences and/or structures. These include linear and conformational epitope prediction methods such as LBtope (Singh et al, 2013), CBTOPE (Ansari and Raghava, 2010), iBCEEL (Manavalan et al, 2018), iLBE (Hasan et al, 2020) as well as a method that deals with prediction of antibody specific B-cell epitopes (Jespersen et al, 2019)
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
