The assignment of functions to proteins is a bottleneck due to the need of costly and time-consuming molecular experiments. This is the reason why more often data analysis methods are used for protein an- notation. In this thesis I consider an approach based on Deep Learning architectures.
