OPTIMIZING DOPED LIBRARIES BY USING GENETIC ALGORITHMS

摘要: The insertion of random sequences into protein-encoding genes in combination with biological selection techniques has become a valuable tool the design molecules that have useful and possibly novel properties. By employing highly effective screening protocols, functional unique structure had not been anticipated can be distinguished among huge collection inactive together represent all possible amino acid combinations. This technique is severely limited by its restriction to library manageable size. One approach for limiting size mutant relies on 'doping schemes', where subsets acids are generated reveal only certain combinations protein sequence. Three mononucleotide mixtures each codon concerned must designed, such resulting codons assembled during chemical gene synthesis desired mixture level translated protein. In this paper we present doping algorithm "reverse translates' three mononucleotides. designed optimally bias these towards choice. combines genetic local optimization strategies based downhill simplex method. Disparate relative representations (and stop codons) within target set generated. Optional weighing factors employed emphasize frequencies their usage, compensate reaction rates different building blocks (synthons) DNA synthesis. effect statistical errors accompany an experimental realization calculated nucleotide simulated. These simulations show robustness optima respect small deviations from values depends concomitant fitness. Furthermore, calculations probe fitness landscape locally allow preliminary assessment structure.