In Silico Prediction of B-Cell Epitopes in the Outer Membrane Proteins of Leptospira interrogans

Abstract

Leptospirosis is a zoonosis of global reach, and the proteins on the outer surface of its causative spirochaete, Leptospira interrogans, are the natural targets for serodiagnosis and for subunit vaccines. Predicting which parts of those proteins an antibody can recognise, before any peptide is made, is attractive because it promises to narrow a long list of candidates cheaply. This narrative review asks how far in silico prediction of B-cell epitopes can be trusted when it is applied to the outer membrane proteins of L. interrogans, and it reports no new predictions. It draws on published work on leptospiral surface biology, on the sequence-based and structure-based tools in common use, and on benchmarking studies of their accuracy. Well-studied antigens such as LipL32, the immunoglobulin-like Lig proteins and OmpL1 show how predicted regions can be set against experimental findings, but they also show the limits: a region can be antigenic without protecting an animal, and an abundant protein can prove dispensable for infection. Linear propensity scales perform only modestly, machine-learning predictors depend on their training sets, and structure-based tools depend on the quality of the model they are given. The review concludes that prediction is best used to rank and prioritise candidates by consensus, never to declare one, and it lists the experimental steps that any predicted epitope still has to survive.