Within the next few years, as sequencing technologies get faster and cheaper, tenths of thousands of newly sequenced genomes will become available (see for example [1, 2]). In contrast, experimental characterisation of genomes’ functional elements, which is expected to lead to breakthroughs in understanding molecular and cellular processes [3], remains in most cases an expensive and painstakingly slow process. In this scenario, computational biology appears as one scalable way to make sense of the vast amount of sequence data that are being deposited (daily) in the public domain. While computational predictions cannot substitute for experiments, they can generate hypotheses to guide targeted experimental validation. In this seminar, I will illustrate ongoing work that aims to shed some light on the “Dark Matter” of the protein sequence universe [4]. Additionally, I will discuss ideas to improve annotation of pathways, protein complexes and organelles in fully sequenced organisms.