Machine learning approaches for prediction of linear B‐cell epitopes on proteins

Abstract

Identification and characterization of antigenic determinants on proteins has received considerable attention utilizing both, experimental as well as computational methods. For computational routines mostly structural as well as physicochemical parameters have been utilized for predicting the antigenic propensity of protein sites. However, the performance of computational routines has been low when compared to experimental alternatives. Here we describe the construction of machine learning based classifiers to enhance the prediction quality for identifying linear B-cell epitopes on proteins. Our approach combines several parameters previously associated with antigenicity, and includes novel parameters based on frequencies of amino acids and amino acid neighborhood propensities. We utilized machine learning algorithms for deriving antigenicity classification functions assigning antigenic propensities to each amino acid of a given protein sequence. We compared the prediction quality of the novel classifiers with respect to established routines for epitope scoring, and tested prediction accuracy on experimental data available for HIV proteins. The major finding is that machine learning classifiers clearly outperform the reference classification systems on the HIV epitope validation set.

Predicting Flexible Length Linear B-Cell Epitopes

“…Previous methods for predicting linear B-cell epitopes (e.g., 15,17,19,18,20 ) have been evaluated on datasets of unique epitopes without applying any homology reduction procedure as a pre-processing step on the data. We showed that performance estimates reported on the basis of such datasets is considerably over-optimistic compared to performance estimates obtained using the homology-reduced datasets.…”

“…Input sequence windows ranging from 10 to 20 amino acids, were tested and the best performance, 66% accuracy, was obtained using a recurrent neural network trained on peptides 16 amino acids in length. In the method of Söllner and Mayer 19 , each epitope is represented using a set of 1487 features extracted from a variety of propensity scales, neighborhood matrices, and respective probability and likelihood values. Of two machine learning methods tested, decision trees and a nearest-neighbor method combined with feature selection, the latter was reported to attain an accuracy of 72% on a data set of 1211 B-cell epitopes and 1211 non-epitopes, using a 5-fold cross validation test 19 .…”

“…In the method of Söllner and Mayer 19 , each epitope is represented using a set of 1487 features extracted from a variety of propensity scales, neighborhood matrices, and respective probability and likelihood values. Of two machine learning methods tested, decision trees and a nearest-neighbor method combined with feature selection, the latter was reported to attain an accuracy of 72% on a data set of 1211 B-cell epitopes and 1211 non-epitopes, using a 5-fold cross validation test 19 . Chen et al 20 observed that certain amino acid pairs (AAPs) tend to occur more frequently in B-cell epitopes than in non-epitope peptides.…”

Predicting Protective Linear B-Cell Epitopes Using Evolutionary Information

“…However, as shown by Blythe and Flower [5], the performance of such methods is only marginally better than that of random guessing. Hence, several methods based on machine learning and statistical approaches have been recently proposed for predicting linear B-cell epitopes [18,30,32,8,31,10,9].…”

Predicting B Cell Epitope Residues With Network Topology Based Amino Acid Indices

“…This indicates that better methods or new amino acid indices are needed for B cell epitope prediction. New methods such as neural networks, hidden Markov models and support vector machines have been applied to B cell epitope prediction very recently [25,[35][36][37]. However, the performance improvements are still limited.…”