Classifying proteins as extracellular using programmatic motifs andgenetic programming

摘要

As newly sequenced proteins are deposited into the world's evergrowing archive of protein sequences, they are typically immediatelytested by various computerized algorithms for clues as to theirbiological structure and function. One question about a new proteininvolves its cellular location-that is, where the protein resides in aliving organism (extracellular, intracellular, etc.). A paper by J.Cedano et al. (1997) reported a human-created five way algorithm forcellular location created using statistical techniques with 76%accuracy. The article describes a two way classification algorithm thatwas evolved using genetic programming with 83% accuracy for determiningwhether a protein is extracellular. Unlike the statistical calculation,the genetically evolved algorithm employs a large and varied arsenal ofcomputational capabilities, including arithmetic functions, conditionaloperations, subroutines, iterations, memory, data structures, setcreating operations, macro definitions, recursion, etc. The geneticallyevolved classification algorithm can be viewed as an extension (which wecall a programmatic motif) of the conventional notion of a proteinmotif. The genetically evolved program constitutes an instance of anevolutionary computation technique producing a solution to a problemthat is competitive with that produced using human intelligence