Informational content of discrete data from parallel test with similar biological objects: selection of key tests

Results from parallel tests with similar biological objects are frequently interrelated. If the data are discrete which is a typical situation in biological mass-screening, the information-theoretical concept of Shannon's entropy can be used to unreval redundancies within such test and to recognize tests with high informational content. This concept was applied to two examples from the screening for fungicidal activity, and the results were compared with activity-activity relationships obtained from pattern recognition and with variance analysis. There was good agreement, and the entropy calculation indeed yielded a very sensitive measure for the informational structure of the data. If the informational structure is known redundant tests may be eliminated. If the evaluation of quantitative structure-activity relationships is attempted it is furthermore possible to restrict the calculations to key tests with high information.