INFERRING DNA STRUCTURES FROM SEGMENTATION DATA

The analysis of DNA structure from restriction enzyme segmentation data has been viewed by molecular geneticists as one of their simpler analysis problems. This paper treats segmentation problems as a case study in the selection between data-driven and model-driven hypothesis formation. The main purpose of this paper is to set forth some useful considerations for selecting a problem solving approach according to the characteristics of a domain. The case study illustrates why an exhaustive model-driven approach, which operates primarily by ruling out all the wrong answers, is a good approach for this domain. A program called GA1 solves segmentation problems using techniques similar to those used in the DENDRAL program. © 1978.