Team:Paris Bettencourt/AMP Evolver
AMP Evolver
in silico design-build-test cycle
AMP Evolver is the function where the input seed sequence will be modified for multiple times to search for a better sequence that may serve as AMP. The framework is built based on the idea of genetic algorithm, which is commonly used in the computer science for global optimization problems [14].
In the AMP Evolver, the input sequence, or so-called “parent”, will be modified by a predefined mutation generator. The simplest mutation generator is random mutagenesis, in which case we mimic the directed evolution in Evolver. However, the mutation generator could actually even more complicated, as a semi-rational design, in which case, we are actually doing a synthetic biology design-build-test cycle in silico.
In our study, we also tested another semi-rational design of the mutation generator. The generator is designed based on previous literature of the observation on charge distribution in natural AMPs [15]. It seems there is a clustering effect of charged amino acids that may be relevant to the efficacy. Therefore we developed an automated mutation generator, to change the charges in random but close enough positions of the original positive charged amino acids.
After, the modified AMPs, will be selected. In our case, we simply select for AMP that makes the bacteria have a lower survival level. This process will iterate for a few generations to finally optimize the AMP.
User is capable to define the generation of iteration, mutation rate per generation, and survivor number after each selection. Besides, any user who are interested in exploring new AMP engineering or design principles, could replace the mutation generator function by a customized function.
The open-source nature makes the AMP Designer also a good platform or environment, to test researcher’s idea of AMP engineering in a simulation.
By compare different mutation generators to the random mutagenesis, we could understand the mechanism of AMP better and also find good strategies to modify peptides that is even not necessarily well-described by Forest model.
Reference
paris-bettencourt-2018@cri-paris.org 
