Sequence-based prediction of olfactory receptor responses

Computational prediction of how strongly an olfactory receptor responds to various odors can help in bridging the widening gap between the large number of receptors that have been sequenced and the small number of experiments measuring their responses. Previous efforts in this area have predicted the responses of a receptor to some odors, using the known responses of the same receptor to other odors. Here we present a method to predict the responses of a receptor without any known responses, by using available data about the responses of other conspecific receptors and their sequences. We applied this method to olfactory receptors in insects Drosophila melanogaster (both adult and larva) and Anopheles gambiae, and to mouse and human olfactory receptors. We found the predictions to be in significant agreement with the experimental measurements. The method also provides clues about the response-determining positions within the receptor sequences.