Gene Expression and Scalable Genetic Search

Abstract

Gene expression evaluates the genetic fitness of an organism through a sequence of representation transformations (DNA→mRNA→Protein). Moreover it does so in a very distributed and decomposed fashion by evaluating different portions of the DNA in order to produce various proteins in different body cells. This chapter reviews some of the recent results that underscore the possible critical role of gene expression in scalable genetic search. It considers a Fourier basis representation to analyze genetic fitness functions and shows that polynomial-time construction of a decomposed representation in the Fourier basis is possible when the function has a polynomial-size description. It also points out that genetic code-like transformations may offer us a unique technique to transform some functions of exponential description in the Fourier basis to an exponentially long representation with only a polynomial number of terms that are exponentially more significant than the rest. This may be useful for a polynomial-time approximation of an exponential description. Since the construction of decomposed representation of functions from observed data plays an important role in machine learning, data mining, and black-box optimization, the role of gene expression in scalable genetic search appears quite critical.