The present invention provides methods and compositions for producing heterologous protein with improved yield and/or quality. A library of randomized ribosomal binding site sequences is provided for the identification of a translation initiation region sequence optimal for expression of the heterologous protein. Also provided are novel ribosomal binding site sequences, and vectors and host cells having those sequences. The library of randomized sequences is useful for screening for improved expression of any protein of interest, including therapeutic proteins, hormones, a growth factors, extracellular receptors or ligands, proteases, kinases, blood proteins, chemokines, cytokines, antibodies and the like.