EPMLR: sequence-based linear B-cell epitope prediction method using multiple linear regression
© Lian et al.; licensee BioMed Central. 2014
Received: 10 July 2014
Accepted: 9 December 2014
Published: 19 December 2014
B-cell epitopes have been studied extensively due to their immunological applications, such as peptide-based vaccine development, antibody production, and disease diagnosis and therapy. Despite several decades of research, the accurate prediction of linear B-cell epitopes has remained a challenging task.
In this work, based on the antigen’s primary sequence information, a novel linear B-cell epitope prediction model was developed using the multiple linear regression (MLR). A 10-fold cross-validation test on a large non-redundant dataset was performed to evaluate the performance of our model. To alleviate the problem caused by the noise of negative dataset, 300 experiments utilizing 300 sub-datasets were performed. We achieved overall sensitivity of 81.8%, precision of 64.1% and area under the receiver operating characteristic curve (AUC) of 0.728.
We have presented a reliable method for the identification of linear B cell epitope using antigen’s primary sequence information. Moreover, a web server EPMLR has been developed for linear B-cell epitope prediction: http://www.bioinfo.tsinghua.edu.cn/epitope/EPMLR/.
The humoral immune response is based on the amazing ability of antibodies to recognize and bind to antigens of intruding organisms, such as bacteria and viruses . Antibodies bind specifically to a contiguous amino acid sequence of a protein known as the linear B-cell epitope or to a folded structure formed by discontinuous amino acids known as the conformational B-cell epitope ,. Prediction of B-cell epitopes is critical for immunological applications. Specifically, predicted peptides can be synthesized and can be used to replace the intact antigen molecules as reagents for detecting anti-protein antibodies in immunoassay , as immunogens for raising anti-peptide antibodies to cross-react with the protein of interest , or in the development of synthetic peptide vaccines . Although the majority of B-cell epitopes are conformational , most B-cell epitopes prediction approaches concentrate on the “easier” linear epitopes .
Earliest linear B cell epitope prediction models were based on propensity profiling. Blythe and Flower  demonstrated that the propensity profiling methods cannot be used to reliably predict the epitope. Even the best propensity profiling method only yielded a success rate marginally better than that produced randomly using a receiver operating characteristics (ROC) plot. Later, machine learning methods have been explored to improve the prediction performance -. However, most of these methods were developed on very small datasets (~872 epitopes and non-epitopes) with negative dataset that were randomly selected peptides instead of experimentally verified non-epitopes .
In this work, based on the antigen’s primary sequence information, a novel linear B-cell epitope prediction model was developed using the multiple linear regression (MLR). A large dataset called BEOD which was derived from BEOracle dataset  was used to train and test our model. It is worthwhile to note that all epitopes and non-epitopes of our BEOD dataset were experimentally verified. Nevertheless, experimental non-epitope data still have the potential to be epitopes due to flawed interpretation of the results or simple experimental errors . Models built on different subsets of such noisy negative dataset may produce very different results. In order to alleviate the noisy problem caused by the negative dataset and report a reliable prediction result of our model, we have performed 300 experiments utilizing 300 sub-datasets of which each negative sub-dataset was randomly selected from the BEOD negative dataset while each positive sub-dataset was the unchanged BEOD positive dataset. 10-fold cross-validation was employed to evaluate the performance of our model. Our model produced average sensitivity (Sn) of 81.8%, precision (P) of 64.1% and area under the receiver operating characteristic curve (AUC) of 0.728 over the 300 experiments. A web server EPMLR implementing linear B cell epitope prediction is available at: http://www.bioinfo.tsinghua.edu.cn/epitope/EPMLR/.
Sliding window size selection
Summary of the 300 trials’ performances using 10-fold cross-validation
81.8 ± 0.8
64.1 ± 0.2
0.719 ± 0.08
Comparison with Other Prediction Methods
Next, we compared our method with SVMTriP method which is a recently published large dataset based method . We performed a 5-fold cross-validation on the SVMTriP dataset. Our method obtained Sn of 80.56% and P of 54.9% which is similar to the performance of SVMTriP method (Sn of 80.1%, P of 55.2%) using 5-fold cross-validation. Our method observed similar Sn (81.8% vs. 80.56%) but a decreased P (64.1% vs. 54.9%) on the BEOD dataset and SVMTriP dataset. The decreased P value could be resulted from the fact that the negative non-epitope dataset of the SVMTriP dataset was from the remaining segments which have not been marked as epitopes in the corresponding antigen sequences.
Similarly, we compared with LBtope method which is the most recently published large dataset developed method . We applied our method to the Lbtope_Fixed_non_redundant dataset (LFNR) whose epitopes and non-epitopes were all experimentally verified. Using the same experimental procedure of LBtope, on the LFNR dataset, our method obtained an AUC of 0.62, which is comparable to the AUCs (0.57 ~ 0.69) obtained by LBtope method by training using 5-fold cross-validation on 90% of the data and testing on the remaining 10% of the data with various features.
Comparison of EPMLR with other methods
54.38 ~ 65.88
57.31 ~ 63.97
55.85 ~ 64.86
0.57 ~ 0.69
The development of epitope prediction research was accompanied by the development of a large and experimentally well-characterized dataset that comprises both positive epitopes and negative non-epitopes . In contrast to the simplicity of the construction of a positive dataset, the construction of a negative dataset has been still debated. Non-epitopes were not used in the early studies. Some authors attempted to construct negative datasets by randomly choosing peptides either from a protein database (such as Swiss-Prot) where no antibody binding is reported or from the antigen areas not encompassing any of the reported epitopes. In recent years, researchers have begun to construct negative datasets from the immune epitope database IEDB  database. IEDB collects both epitopes and non-epitopes from experimentally validated data. However, experimental non-epitope data still have the potential to be epitopes due to flawed interpretation of the results or simple experimental errors . Thus, models built on different subsets of such uncertain dataset may produce uncertain predictions, as demonstrated by the results of the 300 trials of our model. Although we can produce a good result by subjectively selecting a self-reinforcing negative dataset, the reliability of such good performance is not guaranteed. Thus, in this work, we performed many parallel trials using the same positive dataset but different negative datasets that are randomly selected from the noisy negative dataset and reported the average of all results as the final result. Such an averaging method could help produce a reliable result.
In this work, a novel sequence-based linear B cell epitope prediction model was developed. A web server EPMLR implementing the prediction is available at: http://www.bioinfo.tsinghua.edu.cn/epitope/EPMLR/. As a reliable method developed based on a large dataset, EPMLR offers new insights into the linear B cell epitope prediction and a new option for scientists to do their prediction.
In this work, we used BEOracle dataset because it is a large dataset and both epitopes and non-epitopes of BEOracle dataset were experimentally verified. Through combining entries from IEDB, BCIPEP and AntiJen databases, Wang and his colleagues constructed the BEOracle dataset . They extended these epitope sequences equally on both sides to get epitopes of a final length of 100 amino acids using the Uniprot identifiers associated with them. Further, we trimmed BEOracle dataset (100-mer) from both ends equally to extract the core 20-mer peptides. Finally, we obtained 4,405 epitopes and 8,467 non-epitopes and we called this unbalanced BEOracle-Derived dataset (BEOD).
To alleviate the problem caused by the noise of negative dataset, we then constructed 300 sub-datasets which were the same in the positive dataset but different in the negative dataset. Each of the 300 sub-datasets contained the whole 4,405 epitopes of BEOD and an equal number of 4,405 non-epitopes randomly selected from the 8,467 BEOD non-epitopes. These 300 sub-datasets were used to perform the 300 experiments using the same algorithm.
The SVMTriP dataset, which was introduced by Yao B et al. , consists of 4925 epitopes and 4925 non-epitopes. Originally, total of 65,456 B-cell linear epitopes were downloaded from IEDB (version June 11th, 2012) and the identical epitopes and those possibly related to T-cell are removed. Next, truncation and extension technique was applied to get fixed length pattern. Finally, 4925 non-redundant epitope sequences were obtained after > = 30% similarity process by BLAST . For the negative dataset, the same number of equal-length sub-sequences were extracted from the non-epitopic segments in the corresponding antigen sequences.
The Lbtope_Fixed_non_redundant dataset (LFNR), which was introduced by Singh H et al. , consists of 7824 B-cell epitopes and 7853 non-epitopes. Originally, total of experimentally validated 49694 B-cell epitopes and 50324 non B-cell epitopes were obtained from the IEDB in Jan 2012. After truncation and extension, sequences with fixed length were created. Then identical epitopes and common patterns in both types of patterns were removed. Finally, after 80% non-redundant process by CD-HIT , 7824 B-cell epitopes and 7853 non-epitopes were kept. This non-redundant and fixed length dataset was named Lbtope_Fixed_non_redundant.
In this study, we constructed an epitope prediction model based on primary sequence information. The modeling trial was performed as follows.
Here, subscripts j and k denote position j and k in the window. R j is a 19-D vector with the component for the residue at position j as 1 and the others as 0. α(1, 2 … 19|ω) is the coefficient vector for 19 amino acids (with one omitted). represent the features of occurrence of amino acid type from the first position to the last position for an n-mer window sequence. B j and B k are the normalized hydrophilicity values of residues at positions j and k, while β j,k(ω) is the coefficient combining the residue pair. represent the features of autocorrelation of the hydrophobicity index of residue pair (residue R j at position j and residue R k at position k) for an n-mer window sequence. Similarly, S j and S k are the normalized side chain mass values of residues at positions j and k, while γ j,k(ω) is the coefficient combining the residue pair. represent the features of autocorrelation of the side chain mass of residue pair for an n-mer window sequence. V(R j R k ) is a 500-D vector whose components refer to 500 most important position specific residue pairs R j R k , while δ j,k(ω) is the coefficient combining the residue pair. In model training, we compared the 500 R j R k with all R j R k (n × (n − 1)/2 in total) existed in a window, the value of a component of V(R j R k ) is set as 1 if the R j R k to which the component referred exists in the window, otherwise as 0. represent the feature of occurrence of selected residue pairs in an n-mer window sequence.
where f t (R j R k ) represents the occurrence frequency of a R j R k derived from the epitope (t = 1) and non-epitope (t = 0) in the training dataset, respectively. P t represents the naturally occurring probability of a R j R k based on the relative sizes of the epitope and non-epitope datasets in the training dataset (for example, P 1 = P 2 = 0.5 if the size of the epitope dataset is equal to the size of the non-epitope dataset). All D(R j R k ) values were ranked by the descending orders. Finally, 500 R j R k with the largest value of D(R j R k ) were selected. Here, we selected 500 components because the curve of all D(R j R k ) values by descending order shows as exponential decay and the point of inflection is about 500 ( Additional file 1).
On the training dataset, all the fitting coefficients in Equation (1) were determined by the MLR method . Once the coefficient matrix is obtained, we adopted the same sliding window procedure with the 20-mer peptides on the testing dataset. Each of the n-sized window ω i of the 20-mer peptide was predicted to be an epitope or not with an epitope propensity score Q(ω i ). For any 20-mer peptide, there are 21 − n windows and the epitope propensity score of the 20-mer peptide was calculated by taking the average of all 21 − n Q(ω i ) scores. In this representation, every 20-mer peptide in the testing dataset is scored for its propensity to be an epitope or a non-epitope.
In 10-fold cross-validation test, the original dataset is randomly partitioned into 10 equal size subsets. Of the 10 subsets, a single subset is retained as the validation data for testing the model, and the remaining 10-1 subsets are used as training data. The cross-validation process is then repeated 10 times, with each of the 10 subsets used exactly once as the validation data. The 10 results can then be averaged to produce a single estimation.
where TP, TN, FP, and FN represent the number of true positive, true negative, false positive, and false negative cases, respectively.
This work was supported by grant 2009CB918801 from the Ministry of Science and Technology of China. This work was also supported by grant from the Natural Science Foundation of China (No. 31370855). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
- Getzoff ED, Tainer JA, Lerner RA, Geysen HM: The chemistry and mechanism of antibody binding to protein antigens. Adv Immunol. 1988, 43: 1-98. 10.1016/S0065-2776(08)60363-6.View ArticlePubMedGoogle Scholar
- Barlow DJ, Edwards MS, Thornton JM: Continuous and discontinuous protein antigenic determinants. Nature. 1986, 322 (6081): 747-748. 10.1038/322747a0.View ArticlePubMedGoogle Scholar
- Caoili SE: Hybrid methods for B-cell epitope prediction. Methods Mol Biol. 2014, 1184: 245-283. 10.1007/978-1-4939-1115-8_14.View ArticlePubMedGoogle Scholar
- Leinikki P, Lehtinen M, Hyoty H, Parkkonen P, Kantanen ML, Hakulinen J: Synthetic Peptides as Diagnostic-Tools in Virology. Adv Virus Res. 1993, 42: 149-186. 10.1016/S0065-3527(08)60085-8.View ArticlePubMedGoogle Scholar
- Van Regenmortel MHV: Immunoinformatics may lead to a reappraisal of the nature of B cell epitopes and of the feasibility of synthetic peptide vaccines. J Mol Recognit. 2006, 19 (3): 183-187. 10.1002/jmr.768.View ArticlePubMedGoogle Scholar
- Yadav M, Liebau E, Haldar C, Rathaur S: Identification of major antigenic peptide of filarial glutathione-S-transferase. Vaccine. 2011, 29 (6): 1297-1303. 10.1016/j.vaccine.2010.11.078.View ArticlePubMedGoogle Scholar
- Pellequer JL, Westhof E, Vanregenmortel MHV: Predicting Location of Continuous Epitopes in Proteins from Their Primary Structures. Method Enzymol. 1991, 203: 176-201. 10.1016/0076-6879(91)03010-E.View ArticleGoogle Scholar
- Flower DR: Immunoinformatics and the in silico prediction of immunogenicity. An introduction. Methods Mol Biol. 2007, 409: 1-15. 10.1007/978-1-60327-118-9_1.View ArticlePubMedGoogle Scholar
- Blythe MJ, Flower DR: Benchmarking B cell epitope prediction: underperformance of existing methods. Protein Sci. 2005, 14 (1): 246-248. 10.1110/ps.041059505.View ArticlePubMed CentralPubMedGoogle Scholar
- Saha S, Raghava GPS: Prediction of continuous B-cell epitopes in an antigen using recurrent neural network. Proteins. 2006, 65 (1): 40-48. 10.1002/prot.21078.View ArticlePubMedGoogle Scholar
- Chen J, Liu H, Yang J, Chou KC: Prediction of linear B-cell epitopes using amino acid pair antigenicity scale. Amino Acids. 2007, 33 (3): 423-428. 10.1007/s00726-006-0485-9.View ArticlePubMedGoogle Scholar
- Wee LJK, Simarmata D, Kam YW, Ng LFP, Tong JC: SVM-based prediction of linear B-cell epitopes using Bayes Feature Extraction. BMC Genomics 2010, 11(Suppl 4):S21.,Google Scholar
- El-Manzalawy Y, Dobbs D, Honavar V: Predicting linear B-cell epitopes using string kernels. J Mol Recognit. 2008, 21 (4): 243-255. 10.1002/jmr.893.View ArticlePubMed CentralPubMedGoogle Scholar
- Gao JZ, Faraggi E, Zhou YQ, Ruan JS, Kurgan L: BEST: Improved Prediction of B-Cell Epitopes from Antigen Sequences. Plos One 2012, 7(6):e40104.,Google Scholar
- Sollner J, Mayer B: Machine learning approaches for prediction of linear B-cell epitopes on proteins. J Mol Recognit. 2006, 19 (3): 200-208. 10.1002/jmr.771.View ArticlePubMedGoogle Scholar
- Sweredoski MJ, Baldi P: COBEpro: a novel system for predicting continuous B-cell epitopes. Protein Eng Des Sel. 2009, 22 (3): 113-120. 10.1093/protein/gzn075.View ArticlePubMed CentralPubMedGoogle Scholar
- Rubinstein ND, Mayrose I, Martz E, Pupko T: Epitopia: a web-server for predicting B-cell epitopes. BMC Bioinformatics 2009, 10:287.,Google Scholar
- Rubinstein ND, Mayrose I, Pupko T: A machine-learning approach for predicting B-cell epitopes. Mol Immunol. 2009, 46 (5): 840-847. 10.1016/j.molimm.2008.09.009.View ArticlePubMedGoogle Scholar
- Larsen JE, Lund O, Nielsen M: Improved method for predicting linear B-cell epitopes. Immunome Res. 2006, 2: 2-10.1186/1745-7580-2-2.View ArticlePubMed CentralPubMedGoogle Scholar
- Wang Y, Wu W, Negre NN, White KP, Li C, Shah PK: Determinants of antigenicity and specificity in immune response for protein sequences. BMC Bioinformatics. 2011, 12: 251-10.1186/1471-2105-12-251.View ArticlePubMed CentralPubMedGoogle Scholar
- Yao B, Zhang L, Liang SD, Zhang C: SVMTriP: A Method to Predict Antigenic Epitopes Using Support Vector Machine to Integrate Tri-Peptide Similarity and Propensity. Plos One 2012, 7(9):e45152.,Google Scholar
- EL-M Y, Honavar V: Building classifier ensembles for B-cell epitope prediction. Methods Mol Biol. 2014, 1184: 285-294. 10.1007/978-1-4939-1115-8_15.View ArticleGoogle Scholar
- Wang HW, Pai TW: Machine learning-based methods for prediction of linear B-cell epitopes. Methods Mol Biol. 2014, 1184: 217-236. 10.1007/978-1-4939-1115-8_12.View ArticlePubMedGoogle Scholar
- Goodswen SJ, Kennedy PJ, Ellis JT: A guide to in silico vaccine discovery for eukaryotic pathogens. Briefings Bioinformatics. 2013, 14 (6): 753-774. 10.1093/bib/bbs066.View ArticleGoogle Scholar
- Singh H, Ansari HR, Raghava GPS: Improved Method for Linear B-Cell Epitope Prediction Using Antigen's Primary Sequence. Plos One 2013, 8(5):e62216.,Google Scholar
- Greenbaum JA, Andersen PH, Blythe M, Bui HH, Cachau RE, Crowe J, Davies M, Kolaskar AS, Lund O, Morrison S, Mumey B, Ofran Y, Pellequer JL, Pinilla C, Ponomarenko JV, Raghava GPS, van Regenmortel MHV, Roggen EL, Sette A, Schlessinger A, Sollner J, Zand M, Peters B: Towards a consensus on datasets and evaluation metrics for developing B-cell epitope prediction tools. J Mol Recognit. 2007, 20 (2): 75-82. 10.1002/jmr.815.View ArticlePubMedGoogle Scholar
- Vita R, Zarebski L, Greenbaum JA, Emami H, Hoof I, Salimi N, Damle R, Sette A, Peters B: The immune epitope database 2.0. Nuc Acids Res. 2010, 38 (Database issue): D854-D862. 10.1093/nar/gkp1004.View ArticleGoogle Scholar
- Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nuc Acids Res. 1997, 25 (17): 3389-3402. 10.1093/nar/25.17.3389.View ArticleGoogle Scholar
- Li W, Godzik A: Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics. 2006, 22 (13): 1658-1659. 10.1093/bioinformatics/btl158.View ArticlePubMedGoogle Scholar
- Pan XM: Multiple linear regression for protein secondary structure prediction. Proteins. 2001, 43 (3): 256-259. 10.1002/prot.1036.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.