Ab Initio Prediction of Protein Localization

Just as in real estate, in an organism the location of a protein is important. Knowing the location provides a basis for deciding how to test for functionality. The proteins a specific cell makes give it a special character. Similarly, different sub-cellular compartments do different functions in the cell. We have experimental clues about function for fewer than 40,000 or the currently known 1.2 million proteins. The tools developed in this project will enable the predication of the localization of proteins by considering its structure, especially its surface components which should aid in determining function. The first task will be to develop methods using neural networks to predict localization for proteins of know structure. The method will be generalized to proteins of unknown structure by predicating surface residues, derivative of a method already in hand. A database of know sequence-motifs related to location will be developed and integrated into the prediction method. An atlas of localization will help map knowledge about function over the lifetime of any protein. Localization is an excellent starting point for assigning function to proteins. The atlas is also a vehicle for teaching both molecular biology and informatics. The atlas and tools will be available from servers as developed.