| The goal of proteomics is the complete characterization of all proteins. Efforts to characterize subcellular location have been limited to assigning proteins to general categories of organelles. We have previously designed numerical features to describe location patterns in microscope images and developed automated classifiers that distinguished major subcellular patterns with a high accuracy (including patterns not distinguishable by visual examination).; The research described in this thesis focused on improving the performance of automated interpretation tools for protein subcellular location patterns in several areas: implementing and optimizing new 3D feature, introducing automated approaches for selecting the optimal feature set to describe images for a given collection of proteins and construct an effective partitioning of the proteins by location, identifying proper feature transformation among different image datasets, and studying the effects of imaging resolutions on 3D automated image interpretation.; New features, including edge features and Haralick texture features were implemented and optimized for 3D images, which yielded a near optimal classification accuracy (98%) when combined with DNA independent morphological features. Clustering approaches that group proteins by location were implemented and validated using a limited randomly tagged protein dataset. Objective partitioning of proteins from heterogeneous sources based on their location patterns were achieved by a simple feature transformation procedure. Effects of 3D image resolution on both classification and clustering were evaluated, suggesting the existence of 5 consistent location pattern clusters at various resolutions tested. Finally, the value of automated interpretation tools was demonstrated by the application of these tools in identifying residues responsible for uncovering enzyme TGN exit function. |
