The machine learning approach to problems from computational biology is to learn models from labeled training examples. A common problem in supervised learning is the lack of sufficient training data. Often, however, there exists a large body of data on a different, but related problem. In order to boost performance on our problem of interest, we would like to utilize information from this related domain. The field of domain adaptation formalizes this scenario and provides intelligent algorithms to transfer information between domains. The aim of this work is to compare, extend and develop domain adaptation algorithms suited for problems in computational biology. A central part of this work is the evaluation of domain adaptation methods in large scale computational experiments. To our knowledge, this work constitutes the first thorough experimental comparison of available domain adaptation algorithms in a well controlled experimental framework. Furthermore, we are one of the first to apply such methods to problems in computational biology, which appears to provide a wealth of potential applications. Representative for the class of sequence-based classification problems we chose the problem of splice site prediction - an important subproblem in computational gene finding. In splice site prediction, challenges arise from an extremely high dimensional feature space, unbalanced data sets and the fact that there is very little labeled data available for newly sequenced organisms. To improve performance on the organism of interest, we turn to other organisms as additional sources of information. Experiments were conducted for a set of target organisms of varying phylogenetic distance to the source organism and for multiple data set sizes. Comparing the methods over a range of conditions allows us to draw conclusions under what circumstances which methods perform well. We observed that the methods that learn solutions on all data sources simultaneously perform best. Furthermore, methods based on simple convex combinations provide the most cost-effective solution to the domain adaptation problem. The largest improvements over baseline methods can be observed if the organisms are only distantly related. Using the example of splice site prediction, we can show that domain adaptation algorithms are well suited to considerably boost predictive performance on sequence based classification problems.