vorgelegt dem Rat

iii After realizing its frequency in the last decade, alternative splicing has attracted consid-erable attention. Although several biological phenomena can be explained by alternative splicing today, research has just started to uncover all of its aspects. This thesis investigates three aspects of alternative splicing, mostly by means of computational large-scale analyses. In the first part, we introduce a new approach to predict alternative splicing without using expressed sequence information. Given that our knowledge about the human transcriptome is still incomplete, ab initio prediction of alternative splicing is a rather recent but important research area. In contrast to existing methods, our approach is independent of orthologous sequences, thus it is applicable to a single genome. By introducing an efficient dynamic programming algorithm, we reduce the computational complexity during the search for new splice events compared to a naive algorithm. The use of this algorithm is demonstrated in a genome-wide application, where we predict and verify novel human splice events