A unified approach to the computation and analysis of strings of gene expression states

The problem of deciding whether or not a concentration of messenger RNA corresponds to a gene in expressed state is crucial in the Boolean representation of gene regulatory networks. Since there is no independent mathematical description of gene expressions, the computational answers of this problem are affected by uncertainties in the model used for the design of the computer algorithms, and by the number of concentration measurements available. In this article, we handle these uncertainties by computing their probability distributions and unifying heterogeneous algorithmic models with aggregation rules. This approach renders an algorithmic framework that computes a set of probabilistic strings of states and, within it, the elected string, which results from the collective decision of the algorithms on the state of the gene in each of the concentration measurements. The set of probabilistic strings is endowed with a metric and a pseudo metric function that makes it possible to analyze the elected string against the expected and the highest probability strings in the set, among other considerations. The method is applied on biological data sets and the results are compared with those of four previously published algorithms.