Computational Experiments in Molecular Biology: Searching for the 'Big Picture'

According to the American Heritage Dictiofiary, one of the meanings of ‘experiment’ is: “ . . . a test under controlled conditions that is made to demonstrate a known truth, examine the validity of a hypothesis, or determine the efficacy of something previously untried.” In the light of this general definition we can think of a computational experiment as one in which the object of testing and the controlled conditions are given in symbolic form which can then. w manipulated according to some (also symbolic) rules belonging to a system of rules. This brings us to the observation that computational experiments are always done on models of reality and thereby pertain to the reality itself only as well as the model does. Ergo, if we have an inadequate model, computational experiments will contribute to an inadequate description of the modeled phenomenon (we may recall the proverb ‘garbage in, garbage out’ at this point). In this note we shall ignore the formidable technical difficulties inherent in judging the material adequacy of models and we shall assume that our symbolic representations are adequate. In molecular biology this assumption appears to be safe so long as we focus on nucleic acids and proteins represented by their nucleotide or amino acid sequences. Difficulties may arise if we decide to use secondary, tertiary or quaternary structure representations because the available data are not nearly as reliable as the sequence data are (even in the presence of sequencing errors). We can (roughly) distinguish at least three kinds of computational experiments: