We are only starting to understand the functional networks of the proteins encoded in the now sequenced human genome. The ability to predict gene and protein function from the sequence and network context is linked to this goal.
Protein function may, in some cases, be predicted from protein structure or homology to sequentially similar proteins. In many cases, however, this generates insufficiently precise predictions. In order to infer gene or protein function, many different bioinformatics methods may be employed. At Stockholm University, researchers are using computational approaches, such as hidden Markov models, evolutionary models, and discriminant statistical methods - aiming to predict function, both with regards to proteins, biochemical activity and pathways.