MELEPS: Multiple Expert Linear Epitope Prediction System

Abstract

Taiwan is one of the most important fishery countries in the world due to its leading fishery breeding and farming technologies. However, high-density fishery farming environments are vulnerable to bacteria or viruses and would cause serious losses. Predicting epitope binding segments from pathogenic bacteria is the first step for vaccine and drug development, and bioinformatics technologies could provide effective approaches to facilitate effective prediction of epitope segments. This study integrated six linear epitope prediction systems for prediction of highly antigenic segments through a weighted voting mechanism. Testing datasets were retrieved from Bcipep and IEDB to evaluate the performance of the proposed prediction model. The experimental results showed that the proposed multi-expert system performed better than the six individual prediction system in general. The F1-Score of the proposed system could achieve 69.40% and 60.54% respectively, while the average F1-Scores of the other six systems could only achieve 55.21% and 41.94%. The proposed multi-expert recommendation system outperforms individual linear epitope prediction systems.