Computer algorithms based on pattern recognition are being used in many areas of science and technology to assist the scientist in solving complex, time-consuming, and often tedious real-world problems. The basic premise is to train a computer to efficiently identify a known pattern in an unknown dataset. This needle-in-a-haystack approach is being used in the area of genomics, where there are already several examples of very powerful computational pattern recognition approaches available for searching new sequences for structural motifs, similarities to other proteins and DNA, and predicting secondary structure, based solely on the DNA or amino acid sequence. We believe that macromolecular crystallography can also benefit from the application of pattern recognition to the often daunting task of fitting atoms into an electron density map. The fact that electron density maps are three-dimensional images provides an additional challenge to this technology in that the procedures we are developing in order to find matching patterns must be rotation invariant. To test the validity of our hypothesis we will complete the following aims: 1) we will develop a set of rotation invariant features that can characterize the patterns in regions of an electron density map, 2) we will determine the optimal size of feature regions and the size and type of structural database required to find similar regions of electron density capable of accurately determining structures, and 3) we will develop a methodology to synthesize matched regions to produce coherent local and global models of protein structure. If these goals can be met, we will investigate the feasibility of incorporating knowledge-based methods, neural networks, and other AI techniques to augment the interpretation of structures from electron density maps. In addition, we will attempt to extend this methodology to produce initial structures for electron density maps that are either of poor quality and/or low resolution.

National Institute of Health (NIH)
National Institute of General Medical Sciences (NIGMS)
Exploratory/Developmental Grants (R21)
Project #
Application #
Study Section
Molecular and Cellular Biophysics Study Section (BBCA)
Program Officer
Lewis, Catherine D
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Texas Engineering Experiment Station
Engineering (All Types)
Schools of Engineering
College Station
United States
Zip Code
Ioerger, Thomas R; Sacchettini, James C (2002) Automatic modeling of protein backbones in electron-density maps via prediction of Calpha coordinates. Acta Crystallogr D Biol Crystallogr 58:2043-54
Holton, T; Ioerger, T R; Christopher, J A et al. (2000) Determining protein structure from electron-density maps using pattern matching. Acta Crystallogr D Biol Crystallogr 56:722-34