Predicting antigen evolution

Abstract

Evolution does not happen at random and one strain of a bacterium cannot evolve directly to another arbitrary strain. Nature allows only certain paths in going from one strain to another. Based on this idea, this research aims to identify what the fHbp (factor H binding protein) antigen of the Neisseria meningitidis bacterium is likely to mutate into so that vaccines and therapies can be developed in advance for the most likely mutants. Neisseria meningitidis is the bacterium that causes the potentially devastating disease meningococcal meningitis in humans.

The target of this research was to generate valid new variants of fHbp based on the characteristics of the existing fHbp sequences. The characteristics were that the sequences had specific invariant regions which flanked restricted variable regions, the mutations in each position of the variable regions were highly constrained, there were peptides in the homologous sequences which were correlated or were co-occurring with each other, and there were regions which were more or less likely to mutate than by chance. As part of determining the characteristics, tertiary structures of the existing sequences were also predicted and energy values of those structures were determined. A pipeline of programs was written to generate variants of the fHbp sequence which satisfied these characteristics. The new variants were studied, their tertiary structures were predicted, and energy values of those structures were determined, similar to what was done for the existing variants. New variants whose associated energy values fell into the range defined by existing sequences were deemed to be "allowable" by nature.

Unfortunately, all of the variants of fHbp generated were valid according to our energy criterion. All generated variant being "allowed" is an unlikely result. Therefore, we conclude that a more stringent methodology for evaluating the viability of fHbp variants is necessary.

Another contribution of our work is the program for generating variants. Like the methodology, it is highly modular, and can be easily used as starting platform for research into additional filtering criteria.