| Research Abstract |
Adleman's study of DNA computer indicated that (he information carried by DNA molecules can be processed by a defined mathematical algorithm, which is implemented using molecular biology techniques (Adleman, 1994). His way of solving a combinatorial problem involves the generation of a library of solution candidates and the search for the true solution. The solution was actually found by a rigorous implementation of a computing algorithm, not by trial and error.
DNA molecules in nature carry the blueprints (genes) of proteins. It is known that some mutant genes (the genes with base substitutions or mutations in their base sequence) produce mutant proteins with an enhanced activity or an altered substrate specificity. In molecular biology, such desired mutant genes have been obtained by exploring an artificially generated library of mutants. This search process could be regarded as a computation of an optimization problem by using biological or "wet" techniques (The fitness function for protein, computable on electronic computers, is not known yet). However, this process has been implemented largely by trial and error, rather than according to an appropriate algorithm.
In order to design such a search algorithm for "evolutionary protein engineering", we could take advantage of the empirical knowledge and conventional tools (computer experiments, for example) of computer scientists, if protein evolution is actually an instance of optimization problem, and the desired mutant genes can be assumed to be optima or sub-optima on the fitness landscape. In the present study, we first surveyed the existing search strategies used in computer science, before proposing a possible algorithm for protein engineering. The performance of this algorithm was examined by computer experiments, and a wet method for one-point crossover was then developed for a future Implementation of the Whole search process in a wet manner.
|
