The database of 200 million protein-structure predictions now includes homodimers, adding new biological relevance.