Optimal codon pair bias design (extended abstract)

Abstract

Codon pair bias is the species-specific phenomenon that pairs of adjacent codons appear in genomes with frequencies different than would be predicted under an independence assumption, and thus is indicative of evolutionary selection. The synthetic attenuated virus engineering (SAVE) paradigm to design vaccines creates weak viruses by designing coding sequences that favor underrepresented codon pairs. Designing genes which achieve the absolute minimum codon pair bias with an arbitrary codon distribution is computationally difficult. In this paper, we develop new algorithms for constructing provably optimal codon pair designs under coding constraints of up to 186 codons in under one minute. We explore a variety of search mechanisms, lower bounds, and pruning strategies to optimize sequences. Our results make it possible for the first time to truly evaluate the performance of commonly used design methods, and quantify the potential improvement possible through better algorithms.