Coping with Uncertainty in Temporal Gene Expressions Using Symbolic Representations

DNA microarrays can provide information about the expression levels of thousands of genes, however these measurements are affected by errors and noise; moreover biological processes develop in very different time scales. A way to cope with these uncertain data is to represent expression level signals in a symbolic way and to adapt sub-string matching algorithms (such as the Longest Common Subsequence) for reconstructing the underlying regulatory network. In this work a first simple task of deciding the regulation direction given a set of correlated genes is studied. As a validation test, the approach is applied to four biological datasets composed of Yeast cell-cycle regulated genes under different synchronization methods.