Learning Chomsky-like Grammars for Biological Sequence Families

This paper presents a new method of measur ing performance when positives are rare and investigates whether Chomsky like grammar representations are useful for learning accu rate comprehensible predictors of members of biological sequence families The positive only learning framework of the Inductive Logic Programming ILP system CProgol is used to generate a grammar for recognis ing a class of proteins known as human neu ropeptide precursors NPPs As far as these authors are aware this is both the rst bi ological grammar learnt using ILP and the rst real world scienti c application of the positive only learning framework of CPro gol Performance is measured using both predictive accuracy and a new cost func tion Relative Advantage RA The RA re sults show that searching for NPPs by using our best NPP predictor as a lter is more than times more e cient than randomly selecting proteins for synthesis and testing them for biological activity The highest RA was achieved by a model which includes grammar derived features This RA is sig ni cantly higher than the best RA achieved without the use of the grammar derived fea tures