This paper describes some parallel processing algorithms for the analysis of twodimensional electrophoresis images. One of the largest processor arrays in the world. CLIP4, at University College London, was used in the research. The paper concentrates on algorithms using CLIP4 for querying the data produced on the gel images. It is shown that many typical queries that researchers request can be answered without removing the data from the processor array, eliminating slow transfer times due to the input/output bottleneck problem. The production of these data by CLIP4 has already been described in two previous papers in this journal. Improvements to the system for the future are also presented. ~8 198x Academic Press, Inc.
This paper finishes a trilogy in this journal describing a system for the complete analysis of two-dimensional electrophoresis gel images. The first paper (I) described the CLIP4 machine and algorithms for detection of the spots and comparison between images. In the second paper (2). algorithms for extraction of data, such as area, centroid position, and optical density, were presented. In all algorithms, efficient use was made of CLIP4, one of the largest processor arrays in the world. It was shown that results of analysis could be obtained in a few seconds, not minutes or hours in the case of using serial machines.
This paper presents the final step. How can the biomedical researcher obtain the information he requires once all the data have been extracted? It is shown that a great deal of key queries can be answered in a few steps using the CLIP machine on the data.
One of the traditional problems with parallel computers is that of the l/O bottleneck. The I/O bottleneck is the slow transfer of data between a normal serial computer (the host device) and the processor array itself. Quite often, I/O to the processor takes longer than the processing of the data on the array itself. For example, in this application, we may need to extract the data out of the array for further analysis of the results. It can take longer to do this than the entire analysis processing procedure in the machine. An