In this paper, we propose meta-learning as a general tech- nique to combine the results of multiple learning algorithms, each applied to a set of training data. We detail several meta- learning strategies for combining independently learned c las- sifiers, each computed by different algorithms, to improve overall prediction accuracy. The overall resulting classi - fier is composed of the classifiers generated by the differ- ent learning algorithms and a meta-classifiergenerated by a meta-learning strategy. The strategies described here ar e independent of the learning algorithms used. Preliminary ex - periments using different strategies and learning algorit hms on two molecular biology sequence analysis data sets demon- strate encouraging results. Machine learning techniques a re central to automated knowledge discovery systems and hence our approach can enhance the effectiveness of such systems.