Methods of analysis of protein crystal images

Abstract Several protein crystallization techniques, including the vapor diffusion method, lend themselves well to automation techniques. Up until the present time, automation techniques have been restricted to setting up crystallization experiments, and procedures to monitor and analyze the experiments have not been developed. These procedures require additional hardware for video monitoring of crystallization chambers and automatic recognition of protein crystals. An automated image acquisition and analysis system makes use of both image processing routines and pattern recognition procedures. In order to design and implement such a system, we are presently developing algorithms which can recognize and locate protein crystals in video images of crystallization droplets. Images of crystallization experiments are acquired and digitized, and analyses of the droplet images are conducted on the microcomputer which also acts as a host in our laboratory robotics system. We describe here our current progress in designing the image analysis system, including the development of appropriate pattern recognition methods. In addition, the usefulness of various pattern recognition schemes for monitoring the progress of crystallization is explored.