Vorhersage von Genduplikationen und deren Entwicklung in der Evolution

Evolution proceeds via gene duplication and modification; it is through their successive application that nature has created the vast diversity of current genes. When a gene encoding an essential function is duplicated, one copy must preserve the function for its organism. The other, however, is free to be modified, using its ancestral gene as its starting point. Knowing precisely when ancestral gene duplications occurred is indispensable for reconstructing reliable phylogenetic relationships which in turn allows us to refine the prediction of a genetic function. For some gene families, such as globines or rhodopsines, ancestral gene duplications are well predicted, but in general they are unknown. Mapping out this unknown territory is an important open issue in molecular biology today. Recognizing this, we designed an algorithm that extends a successful but in- formal model built by biologists. This algorithm predicts gene duplications by exploiting inconsistencies in two phylogenetic trees, a gene tree and a species tree. The gene tree represents the phylogeny of a set of current genes, while the species tree represents the phylogeny of the current genes organized by their host-species. As output the algorithm calculates the gene tree reconciled with its predicted gene duplications, which is called a reconciled tree. For all applications that are possible in practice, this algorithm runs in linear time in the size of the given gene tree. This time complexity allows us to screen the today's fast growing molecular databases for gene duplications. First, we partition a genetic database into gene-families and reconstruct a gene tree for each family by one of the well established and fast reconstruction algorithms. For each gene tree, we then find its species tree in a morphologic database. Finally we predict the gene duplications for every gene and species tree pair. Thus, we are able to predict hidden gene duplications throughout the genetic database. While our algorithm predicts gene duplications, it simultaneously associates a reliability index with the set of predictions it has output. Since the reconciled tree is quadratic in the size of the gene tree, we developed a new graphical representation to ease its analysis by biologists. In essence, this graphical representation is the gene tree reconciled only at those gene duplications that are of interest for the biologist. Through this logical compression in presentation, we can, in general, avoid the quadratic explosion. Up to now, this work is of theoretical nature, but its transformation into applications for biologists has begun