EMBL, Heidelberg and MDC, Berlin, Germany
To assign functional features to sequences, various bioinformatics tools are currently used. Most of them have only a limited accuracy so that information is lost in the process. In addition, it is hard to quantify the functional similarity of related sequences, so that even for clearly related sequences the prediction of function is not trivial. As there are different functional levels stored in undigested text format in current sequence databases, the struggle with the (partly even contradictory) annotation is becoming more and more the bottleneck in the annotation process. Apart from reviewing the procedures we use for functional annotation of sequences, I will give some more details on two tools: 1) GIN, a homology- and neural network-based gene prediction program and 2) a program for the identification of orthologues for a given query based on blast postprocessing steps.