How VaxPress Works

Optimization

VaxPress utilizes a genetic algorithm to produce an optimized mRNA CDS sequence. It initiates a collection of randomly mutated child sequences based on the CDS sequence. The number of these sequences is determined by the --population parameter. Subsequently, scoring functions evaluate these sequences. A selection process then identifies the fittest sequences, which are carried forward to the next generation. The number of these survivors is defined by the --survivors parameter.

In the following iteration, these survivor sequences produce new offspring sequences under a pre-defined mutation rate. This process persists for a specified number of iterations, set by the --iterations parameter. Over hundreds to thousands of these cycles, the sequence population progressively evolves towards the optimal coding sequences, exhibiting improved fitness with each iteration.

Objective Function

The objective function is a linear combination of the factors below, with associated weights. It is represented as follows:

\[\textup{maximize}\ Fitness(s) = \Sigma_{f \in factors} weight_{f} \cdot metric_{f}(s)\]

Refer to the Scoring Functions section for more details on each factor.

Adaptive Decrement of Mutation Rate

To achieve the best results, generating new populations that can effectively compete with their predecessors is essential. As the optimization process progresses, new populations with many mutations may not be selected due to their predecessors’ high fitness. If the fitness score remains constant despite ongoing optimization, reducing the mutation rate to enhance the existing population is necessary. The --winddown-trigger parameter is the count of iterations with a constant fitness score that necessitates a decrease in the mutation rate. The --winddown-rate parameter is the factor by which the mutation rate is reduced when the winddown is activated.