Programmatic Moti f s and Genetic Programming

As newly sequenced proteins are deposited into the world's ever-growing archive of protein sequences, they are typically immediately tested by various computerized algorithms for clues as to their biological structure and function. One question about a new protein involves its cellular location – that is, where the protein resides in a living organism (extracellular, intracellular, etc.). A 1997 paper reported a human-created five-way algorithm for cellular location created using statistical techniques with 76% accuracy. This paper describes a two-way classification algorithm that was evolved using genetic programming with 83% accuracy for determining whether a protein is extracellular. Unlike the statistical calculation, the genetically evolved algorithm employs a large and varied arsenal of computational capabilities, including arithmetic functions, conditional operations, subroutines, iterations, memory, data structures, set-creating operations, macro definitions, recursion, etc. The genetically evolved classification algorithm can be viewed as an extension (which we call a programmatic motif) of the conventional notion of a protein motif. The genetically evolved program constitutes an instance of an evolutionary computation technique producing a solution to a problem that is competitive with that produced using human intelligence. 1 . Background and introduction Proteins are responsible for such a wide variety of biological structures and functions that it can be said that the structure and functions of living organisms are primarily determined by proteins (Stryer 1995). Proteins are composed of a chain of amino acid residues in a linear arrangement. The same 20 amino acid residues (denoted by the letters A, C , D, E, F, G , H, I, K, L, M , N, P, Q, R, S, T, V, W , and Y) are used for virtually all proteins of all species on earth. Thus, a protein can be viewed as a linear sequence over a 20-letter alphabet (called the primary structure of the protein). The length of protein sequences vary widely, with the average being about 300. For example, the primary structure of bovine pancreatic trypsin inhibitor (BPTI) contains only 58 amino acid residues, as shown below: RPDFCLEPPY TGPCKARIIR YFYNAKAGLC QTFVYGGCRA KRNNFKSAED 50