Strategies and techniques for testing the precision, reliability and reproducibility of computerized two-dimensional...

A set of test procedures is described for evaluating the reliability, precision and reproducibility of computer systems designed to analyze two-dimensional gel electrophoretograms. The Elsie 4 gel analysis system (Analyt. Biochem., 169, 49-70 (1988) is used to demonstrate the use of these tests. Three major groups of tests are described: analysis of the scanner; analysis of mathematically constructed model gels; and analysis of real gel images. Scanner tests involve evaluating the stability and reproducibility of the scanner. The tests consist primarily of measuring the output of the scanner over a time period to determine its stability, and evaluating the consistency of the scanner at different points in the scanning field. Commonly encountered real gel situations are simulated and analyzed by arranging computer-generated 'ideal spots' in different ways. With such gels we can determine such things as the ability of the computer system to separate closely resolved and/or shoulder spots; whether or not streaks and/or smears are handled correctly; how random noise affects measurements; and, in idealized situation, the accuracy of the quantitation. The idealized spots are generated using the two-dimensional Gaussian as a model of density distribution; other spot models can be used. The Elsie 4 system is capable of finding virtually any spot (if detection parameters are set low enough), and of resolving different-sized spots whose peaks are separated by one and one-half their mean half width). The most significant tests are done on a set of actual gels.(ABSTRACT TRUNCATED AT 250 WORDS)